Keywords: Swarm Intelligence, Advanced Optimization, Decision Making
Abstract: This paper proposes a spider monkey optimization (SMO) based method for optimal operational planning of energy plants. SMO is one of the cooperative metaheuristics. In the optimal operation of energy plants in commercial buildings and factories, it is required to determine startup and shutdown status of each machine and its input/output values during startup, as well as to consider linear and nonlinear machine characteristics. Thus, this problem can be formulated as a mixed integer non-linear optimization problem, and it requires application of evolutionary computation methods. The effectiveness of the proposed spider monkey optimization (SMO) based method is verified by comparing with the conventional differential evolutionary particle swarm optimization (DEEPSO), brain storm optimization (BSO), modified BSO (MBSO), and multi-population MBSO (MP-MBSO) based methods. The results are verified using the Kruskal-Wallis test and the Mann-Whitney U tests with holm correction.

Keywords: Transportation and Vehicle Systems, Swarm Intelligence, Operations Research
Abstract: In the multiple traveling salesman problem (TSP), a group of cities to be visited has been assigned to each salesman based only on the cities’ geographic information, and the visiting routes of the salesmen are planned. However, there is no guarantee that the adopted clustering method is appropriate for route planning. In this study, we proposed a two-stage search method, where the clustering is performed using an artificial neural network, its weights are designed through a multi-objective evolutionary algorithm (MOEA), and each salesman’s visiting route is solved using a TSP solver. We conducted computational experiments for a test problem to compare the performance of the proposed method to a canonical clustering method. Additionally, we examined the characteristics of the balanced solution selected from the obtained non-dominated solution set.

Keywords: Model-Based, Particle Swarm Optimization, Explainability
Abstract: In this research, we introduce a methodology that combines digital twins and Particle Swarm Optimization (PSO) to improve real-time adaptability and interpretability in industrial systems. Using an industrial DC motor simulation as a case study, our approach involves creating a digital twin, performing online parameter estimation via PSO, and identifying unknown system components. The results, especially from scenarios like armature resistance degradation and unbalanced shaft conditions, highlight the digital twin's accuracy and adaptability. This work showcases the potential of our method for real-time monitoring and proactive maintenance in industrial applications.

Keywords: Image Processing
Abstract: Image stitching technology has application scenarios in many fields. This algorithm achieves the acquisition of simultaneously synthesized images, using the RGB module of the Intel Realsense D435 camera for image acquisition. Firstly, a raw image is created to store the final result, and one image is collected at 100ms intervals each time. Images with similarity less than 5/8 are taken and saved. Use the SIFT scale invariant feature detection algorithm to extract image feature points from the collected images, use RANSAC to extract feature points, use the random sample consistency algorithm to filter effective points, and calculate the homography transformation matrix. Synthesize every two collected images into one image and overlay it at the corresponding position in the resulting image. Through the experiment in this article, the average time for single acquisition and synthesis is 70ms, achieving the real-time goal. The similarity between the experimental group and the control group can reach 70%, and the resolution has been increased by 1.65 times, achieving the goal of continuous splicing.

Keywords: Image Processing, Deep Learning, E-health
Abstract: The utilization of generative models in image synthesis has become increasingly prevalent. Synthetic medical imaging data is of paramount importance, primarily because authentic medical imaging data is scarce, costly, and encumbered by legal considerations pertaining to patient confidentiality. Consent forms are typically required from patients in order to utilize their data for publication in medical journals or educational purposes. Consequently, the accessibility of medical data for general public research is limited. Synthetic medical images offer a potential resolution to these issues. The predominant approaches primarily assess the quality of images and the degree of resemblance between these images and the original ones employed for their generation. In this study, we employ a CycleGAN model to produce artificial images depicting several types of pneumonia, including general, bacterial, and viral pneumonia. We then evaluate the performance of these synthetic images by comparing them with ratings made by three respiratory care professionals. Consequently, a range of pneumonia pictures were acquired, exhibiting diverse levels of performance, ranging from being easily identified as false to being correctly identified as real in over 80% of cases.

Keywords: Randomized Algorithms, Pattern Recognition
Abstract: One of the most frequently used models for time series prediction is the Long Short Term Memory (LSTM). LSTM can leverage the past patterns to efficiently forecast the future observations but it is often criticized as very computationally expensive due to the iterative training. In this paper, to reduce the computational workload and improve the prediction performance of time series, we propose a novel auto-regression framework based on Random Vector Functional Link (RVFL). The new framework offers a lighter network structure with higher training efficiency compared to LSTM-based approaches. It is a new attempt to utilize randomized learning algorithms for time series prediction, providing valuable insights for developing faster and more efficient models in the future.

Keywords: Deep Learning, Image Processing, Pattern Recognition
Abstract: High-throughput phenotyping (HTP) of seeds is the comprehensive assessment of complex seed traits and the measurement of parameters that form more complex traits. The key aspect of seed phenotyping is cereal yield estimation. While mechanized seed kernel counters are available in the market currently, they are often priced high and sometimes outside the range of small scale seed production firms’ affordability. The development of object tracking neural network models such as You Only Look Once (YOLO) enables computer scientists to design algorithms that can estimate cereal yield inexpensively. The key bottleneck with neural network models is that they require a plethora of labelled training data before they can be put to task. We demonstrate that the use of synthetic imagery serves as a feasible substitute to train neural networks for object tracking. Furthermore, we propose a seed kernel counter that uses a low-cost mechanical hopper, trained YOLOv8 neural network model, and object tracking algorithms on StrongSORT and ByteTrack to estimate cereal yield from videos. The experiment yields a seed kernel count with an accuracy of 95.2% and 93.2% for Soy and Wheat respectively using the StrongSORT algorithm, and an accuray of 96.8% and 92.4% for Soy and Wheat respectively using the ByteTrack algorithm.

Keywords: Deep Learning
Abstract: The development of Large Language Models (LLMs) like GPT-series and BLOOM has revolutionized Artificial Intelligence, yet it has also brought forth challenges in misuse, such as fake content generation and academic cheating. Detecting whether a text is generated by an LLM or written by a human has become imperative. Fine-tuned LLMs have proven to be a promising approach in this regard. In our study, we fine-tuned seven LLMs (BERT, DeBERTa-v3, RoBERTa, XLMRoBERTa, GPT-2 Medium, GPT-2 Large, GPT-2 XL) to detect text generated by even larger models (GPT-3 and BLOOM) in the AuTexTification task. Among the models, GPT-2 Medium exhibited the best performance in the testing set, achieving an F1-macro score of 0.83272 and an accuracy of 0.83442, surpassing the benchmark’s best-known result.

Keywords: Deep Learning, Explainability, Graph Neural Networks
Abstract: Deep learning models have proven remarkably adept at extracting salient features from raw data, driving state-of-the-art performance across many domains. However, these models suffer from a lack of interpretability; they function as black boxes, obscuring the feature-level support of their predictions. Addressing this problem, our work presents an innovative framework that fuses the power of convolutional layers for feature extraction with the versatility of Graph Neural Networks (GNNs) to model relationships among neuron activations. Our framework operates in two phases: first, it identifies class-oriented neuron activations by analyzing image features, then these activations are encapsulated within a graph structure. The GNN leverages the relationships among these neuron activations to generate a final, interpretable classification. As a result, predictions can be reverse-engineered to pinpoint the specific contributing neurons, thereby enhancing explainability. The proposed model not only matches, but at times exceeds, the accuracy of current leading models, all the while providing transparency via class-specific feature importance. This novel integration of convolutional and graph neural networks offers a significant step towards interpretable and accountable deep learning models.

Keywords: Dimension Reduction, E-health, Ensemble Learning
Abstract: Inflammatory bowel disease, including Crohn’s disease and ulcerative colitis, is a rising global issue. Accurate diagnosis is vital but challenging. This study used the REFS algorithm to identify IBD biomarkers using three mRNA datasets from the GEO repository. The selected genes demonstrated excellent diagnostic accuracy, highlighting the potential of machine learning in advancing IBD research.

Keywords: Financial Engineering, Decision Making, Deep Learning
Abstract: Traditional trading methods often use fixed-interval sampling to capture price changes. In this work, we use an intrinsic time sampling method referred to as directional changes (DC), which reports information whenever there is a significant price change. Tick data from an array of seven FX currency pairs is sampled using the DC framework. We then compare eleven different machine learning (ML) algorithms in a regression task of predicting when the current trend in the market will reverse. These algorithms are: decision tree, random forest, support vector regression, linear regression, stochastic gradient descent regression, kernel ridge regression, elastic net regression, bayesian ridge regression, gradient boosting regression, multilayer perceptron, and long short-term memory neural network. Predicting trend reversal is crucial in trading, as it allows us to anticipate changes in the market and take the relevant actions that are necessary to maximise our returns. After identifying the best ML algorithm for a dataset, we use this prediction as an input of a DC-based trading strategy, and report its performance in terms of return and risk (maximum drawdown). We also benchmark this strategy against four other trading strategies, which include technical analysis and buy and hold. Results over 349 datasets show that the proposed DC-based trading strategy is able to consistently offer high returns at low risk, statistically and significantly outperforming all other benchmarks.

Keywords: Reinforcement Learning, Smart Grid
Abstract: This study introduces a novel two-degree-of-freedom orientation mechanism for photovoltaic panels, utilizing 3D-printed gears and controlled by the DDPG reinforcement learning algorithm. The research highlights the potential for enhanced solar energy capture. The integration of mechanical design with machine learning showcases a promising interdisciplinary approach to renewable energy systems.

Keywords: Deep Learning, Graph Neural Networks, Big Data
Abstract: This paper explores concatenating pre-trained models for simultaneous facial age transformation and face reenactment, emphasizing image quality enhancement. We introduce an identity recognition loss function during age transformation model development to separate identity and age features, optimizing it with a finely-tuned age prediction model. Our research highlights the success of this concatenated training process, especially in remarkable image generation results.

Keywords: Deep Learning, Automated Algorithm, Pattern Recognition
Abstract: Songs convey emotions through melody and lyrics. They capture feelings in small text fragments. Emotions within songs vary: positive, negative, or neutral. This study merged two datasets to create a third, leveraging LLMs for competitive song text classification results.

Keywords: Fault Detection, Deep Learning
Abstract: One-class classification refers to approaches of learning using data from a single class only. In this paper, we propose a deep learning one-class classification method suitable for multimodal data, which relies on two convolutional autoencoders jointly trained to reconstruct the positive input data while obtaining the data representations in the latent space as compact as possible. During inference, the distance of the latent representation of an input to the origin can be used as an anomaly score. Experimental results using a multimodal macroinvertebrate image classification dataset show that the proposed multimodal method yields better results as compared to the unimodal approach. Furthermore, study the effect of different input image sizes, and we investigate how recently proposed feature diversity regularizers affect the performance of our approach. We show that such regularizers improve performance.

Keywords: Automated Algorithm, Deep Learning, Bio-inspired
Abstract: In this work, we propose a novel representation of optimization problem instances using a transformer-based neural architecture trained for the task of problem classification of the 24 problem classes from the Black-box Optimization Benchmarking (BBOB) benchmark. We show that transformer-based methods can be trained to recognize problem classes with accuracies in the range of 70%-80% for different problem dimensions, suggesting the possible application of transformer architectures in acquiring representations for black-box optimization problems.

Keywords: Automated Algorithm, Evolvable Systems, Explainability
Abstract: The primary objective of this study is to develop a suite of computational engines that leverage automation, optimization, and high-performing computing resources to facilitate high-fidelity (HiFi) seismic risk and resilience assessments. In the context of regional-level assessment, HiFi risk simulations are based on the modern performance-based earthquake engineering (PBEE) methodology, which is designed to conduct individualized and explicit loss (e.g., financial loss and functional recovery time) analysis. The methodology is inherently cumbersome and compute-intensive as it relies on building-, site-, and hazard-specific information that requires substantial data preprocessing. At its core, the methodology systematically transforms seismic hazard into quantifiable risk metrics through three major computational modules: 1) probabilistic seismic hazard analysis (PSHA), 2) probabilistic seismic demand analysis (PSDA), and 3) loss analysis using Monte Carlo simulation. The PSHA is a mathematical representation of the seismic hazard that encompasses uncertainties in the size, location, rate of occurrence, and resulting ground motions that a particular site is likely to observe. Subsequently, the PSDA computes the structural response, ideally via nonlinear response history analyses using the ground motion records selected as part of the PSHA. The distribution of the structural response is ultimately used to perform Monte Carlo simulation-based loss assessment. In this study, the economic loss is assessed following the FEMA P-58 guidelines while the time to regain functionality of a building is calculated based on the ATC-138 procedure. The three modules are executed sequentially as each module hinges on the inputs from the preceding one.

Keywords: Cybersecurity, Ethical AI, Automated Algorithm
Abstract: Sensitive personal information is at risk of exposure by the institutions it is shared. Institutions are responsible for preserving the privacy of the personal data they hold, even more so, in the case of sensitive data. ICIS, a model for context-based identification and classification of sensitive personal information, considers the context to identify personal data in unstructured texts of government type documents, regardless the size and type, and then classify each text segment as sensitive personal information, using natural language processing and machine learning techniques. ICIS not only indicates whether a text segment contains sensitive information or not, it also indicates personal data identified in each text segment, their location in the document and whether each text segment is classified as sensitive information. The main contributions of this work are both the identification of personal data and the classification of sensitive information based on the context, and the definition of sensitive personal information, in computational terms.

Keywords: Decision Making, Fuzzy Systems, Operations Research
Abstract: In the context of e-commerce, establishing a stable alliance is crucial in the logistics industry. Consequently, the challenge of ensuring a fair profit allocation arises. In this study, we address the problem of profit allocation for logistics enterprise coalitions with inadequate information and propose a relevant model. To demonstrate the effectiveness of the model, a case study is provided. The results show that our method enhances the multi-party cooperation and serves as an effective tool for the fair and equitable allocation of profits.

Keywords: Decision Making, Financial Engineering, Deep Learning
Abstract: The real estate asset class has captured the attention of billions of global investors due to its ability to generate consistent returns and offer diversification benefits within a mixed-asset portfolio. Prior research has highlighted the advantages of including real estate in portfolio optimization. However, existing studies have primarily focused on historical data when addressing this optimization problem. This paper presents an analysis of the performance of a portfolio that incorporates real estate using price predictions derived from a Long Short-Term Memory (LSTM) model. To provide a comprehensive evaluation, we compare the performance of our portfolio against a benchmark portfolio consisting of stocks and bonds only. To this end, we run a genetic algorithm on the two portfolios. Our findings demonstrate a substantial improvement in the average risk-adjusted return of the portfolio that includes real estate with a magnitude of around 100%, highlighting the substantial value that real estate brings to a diversified portfolio. In this way, we propose a novel approach for showing the benefits of investing in real estate.

Keywords: Decision Making, Explainability, Ethical AI
Abstract: This paper makes a contribution to the CI platform aimed at enhancing the efficiency of student recruitment procedures. Our study entails a comprehensive follow-up audit of this domain, and identifies the key challenges to the integration of equity-conscious practices into the recruitment process. We propose an innovative solution designed to bridge the relevant socio-technological gaps, that is a self-aware recruitment engine. This engine functions within two interconnected conceptual paradigms: machine learning and probabilistic reasoning. To illustrate our approach, we offer a demonstrative example that showcases its practical application.

Keywords: Data Mining, Fault Detection, Pattern Recognition
Abstract: In our quest to incorporate Augmented Reality (AR) into industrial and university laboratories for training purposes, we conducted an analysis of four AR frameworks. Our goal was to develop a portable Starter Kit (SK) and determine the most sustainable option for our project. This kit integrates both hardware and software components, designed to improve an optimize AR functionality on computers, smartphones, and tablets. The research involves four key stages: beginning with NAS configuration, followed by 3D model creation, next is the generation of QR code identifiers, and finally, the development of a cross-platform (C-P) solution.

Keywords: Data Mining, Decision Making
Abstract: As part of an initiative to mitigate transportation congestion of more than 230,000 daily travellers between Mexico City and Mexico State, the Mexican Federal Government, through the Secretariat of Communications and Transportation, began the construction of the Mexico-Toluca Interurban Train in January 2015. In this paper, an exhaustive analysis of the train infrastructure in an event of a high-magnitude earthquake is carried out. For that, a scenario of a large earthquake in which the train moves at high speed transporting passengers is considered. Specifically, we analyse the structure behaviour when it is exposed to an earthquake comparable in magnitude to those experienced on September 19th 1985, in Mexico, and May 22nd 1960, in Chile. Preliminary results confirm the useful of data science techniques for the study, and offer a comprehensive analysis of the train structural integrity under earthquake conditions via simulations conducted using the SAP 2000 software.

Keywords: Signal Processing, Data Mining, Decision Making
Abstract: n/a

Keywords: E-health, Deep Learning, Image Processing
Abstract: Attention Deficit Hyperactivity Disorder (ADHD) is a common neurobehavioral disorder in humans worldwide. While extensive research has focused on machine learning methods for ADHD detection and diagnosis. Most methods rely on high-cost equipment and trained staff for data collection, e.g., Magnetic Resonance Image (MRI) machine and Electroencephalography (EEG) patch. Therefore, low-cost sensors-based easy-to-process methods for ADHD detection by exploiting action and behavior symptoms are required. We present that skeleton-based action recognition has the potential to address the application due to the action-focused nature of ADHD. Hence, this work proposes a novel ADHD detection system with a privacy-mitigating skeleton-based action recognition framework by utilizing our new real multi-modal ADHD dataset. Compared to the conventional methods, the proposed method shows cost efficiency and significant performance improvement. This method also outperforms the conventional methods in accuracy and AUC on the real multi-modal dataset. Furthermore, our proposed method based on simple non-wearable sensors is widely applicable for ADHD screening.

Keywords: Deep Learning, Dimension Reduction, Image Processing
Abstract: Increasing the model capacity is a known approach to enhance the adversarial robustness of deep learning networks. On the other hand, various model compression techniques, including pruning and quantization, can reduce the size of the network while preserving its accuracy. Several recent studies have addressed the relationship between model compression and adversarial robustness, while some experiments have reported contradictory results. This work summarizes available evidence and discusses possible explanations for the observed effects.

Keywords: Deep Learning, Signal Processing, Ensemble Learning
Abstract: This study addresses the limitations of many gesture recognition algorithms, which predominantly employ machine learning-based approaches tailored to specific types of gestures, leaving niche gestures such as musical conducting gestures largely unexplored. To advance the research in musical conducting gesture recognition, we focus on two key aspects: (1) broadening the dataset to encompass various conducting speeds and investigating its impact on performance, and (2) introducing a stacked neural network architecture to explore performance improvements beyond conventional node increase. The study demonstrates that incorporating diverse data significantly enhances performance and that stacking neural network layers yields notable performance gains.

Keywords: Deep Learning, Decision Making, Data Mining
Abstract: We study how to extend the use of the diffusion model to answer the causal question from the observational data under the existence of unmeasured confounders. In Pearl’s framework of using a Directed Acyclic Graph (DAG) to capture the causal intervention, a Diffusion-based Causal Model (DCM) was proposed incorporating the diffusion model to answer the causal questions more accurately, assuming that all of the confounders are observed. However, unmeasured confounders in practice exist, which hinders DCM from being applicable. To alleviate this limitation of DCM, we propose an extended model called Backdoor criterion-based DCM (BDCM), whose idea is rooted in the Backdoor criterion to find the variables in DAG to be included in the decoding process of the diffusion model so that we can extend DCM to the case with unmeasured confounders. Synthetic data experiment demonstrates that our proposed model captures the counterfactual distribution more precisely than DCM under the unmeasured confounders.

Keywords: Deep Learning, Image Processing
Abstract: In this paper, we aim to improve segmentation performance and uncertainty calibration within a fixed computational budget. We propose PyramidEnsembles, which contain members ranging from small over medium and large, to overcome one major problem in applying neural networks to the automotive domain: the trade-off between model performance and overconfidence in the uncertainty predictions. PyramidEnsembles use multiple models of different sizes from the same family in order to combine their strengths: good segmentation performance and well-calibrated uncertainties. We focus our experiments on EfficientNet-based segmentation models applied to the Cityscapes dataset, which is widely used in the field of autonomous driving. We evaluate single models, uniform ensembles (one architecture repeated) and PyramidEnsembles (combination of different model capacities) composed of the EfficientNet model family. Our evaluations show that within the same computational budget, PyramidEnsembles can outperform a single model in terms of segmentation performance while providing better calibrated uncertainties. Scaling over different computational budgets shows that this performance gap increases further. Different uniform ensembles offer a comparable segmentation or uncertainty calibration performance: 3 copies of the EfficientNet-B3 achieve an IoU of 0.7195 while an ensemble of 7 EfficientNet-B0 models yields an Expected Calibration Error (ECE) of 0.0667. One PyramidEnsemble containing an instance of EfficientNet-B0 through B3 is a close second on either metric at 0.7188 IoU and 0.0698 ECE and offers a better trade-off between segmentation performance and uncertainty calibration in this computational budget.

Keywords: Deep Learning, Explainability, Reinforcement Learning
Abstract: In the context of MDPs with high-dimensional states, downstream tasks are predominantly applied on a compressed, low-dimensional representation of the original input space. A variety of learning objectives have therefore been used to attain useful representations. However, these representations usually lack interpretability of the different features. We present a novel approach that is able to disentangle latent features into a controllable and an uncontrollable partition. We illustrate that the resulting partitioned representations are easily interpretable on three types of environments and show that, in a distribution of procedurally generated maze environments, it is feasible to interpretably employ a planning algorithm in the isolated controllable latent partition.

Keywords: Data Mining
Abstract: Concept drift refers to a change in the data distribution affecting the data stream of future samples. Consequently, learning models operating on the data stream might become obsolete, and need costly and difficult adjustments such as retraining or adaptation. Existing methods usually implement a local concept drift adaptation scheme, where either incremental learning of the models is used, or the models are completely retrained when a drift detection mechanism triggers an alarm. This paper proposes an alternative approach in which an unsupervised and model-agnostic concept drift adaptation method at the global level is introduced, based on autoencoders. Specifically, the proposed method aims to ``unlearn'' the concept drift without having to retrain or adapt any of the learning models operating on the data. An extensive experimental evaluation is conducted in two application domains. We consider a realistic water distribution network with more than 30 models in-place, from which we create 200 simulated data sets / scenarios. We further consider an image-related task to demonstrate the effectiveness of our method.

Keywords: Data Mining
Abstract: Stock price prediction has attracted several investors willing to maximize their profits, believing the opportunities to expand their earnings are higher than using conventional financial approaches, such as savings or fixed deposits. Market analysts, traders, and researchers have investigated different techniques such as Bayesian model, Fuzzy classifiers, Artificial Neural Networks (ANN), Support Vector Machines (SVM), etc. to analyze stock markets and make trading decisions. More recently, deep learning models have gained prominence. However, because of the large amount of data required for training, these techniques typically aggregate all stocks in a single database, creating a generic model. On the contrary, we propose to predict stock price movements considering each stock as a distinct dataset, training specialized machine learning traditional classifiers for each one. We compare the proposed procedure, using different learners, mainly with state-of-the-art deep learning techniques. The results suggest that using specific models for each stock, employing simple and small feature sets, significantly contributes to improved model performance. Our best model, using Logistic Regression, outperformed all the other models.

Keywords: Data Mining
Abstract: Data streams, due to their dynamic nature, tend to impose a number of constraints on the functioning of the learning models used to extract knowledge from these environments. In this context, concept drift is an emerging research area, as they negatively affect the performance of classifiers: after they have been trained with a specific concept, they tend to lose accuracy in the presence of a new concept. Additionally, this problem is often worsened in environments with imbalanced classes, because the identification of changes in the distributions of examples belonging to minority classes is usually more complex, due to their lack of representativeness in the data stream. This work proposes the Features and Classes Drift Detector (FCDD), a new method specially designed to deal with the problem of concept drifts in imbalanced data streams, aiming to maintain the accuracy of detections in minority classes and, in addition, to avoid discarding the knowledge inherent to the classes unaffected by drifts. Experiments conducted in an imbalanced scenario with partial concept drift demonstrated the effectiveness of the proposed method when compared to the current state of the art detectors.

Keywords: Data Mining, Deep Learning
Abstract: Multi-variate time series forecasting plays a crucial role in addressing key tasks across various domains, such as early warning, pre-planning, resource scheduling, and other critical tasks. Thus, accurate multi-variate time series forecasting is of significant importance in guiding practical applications and facilitating these essential tasks. Recently, Transformer-based multi-variate time series forecasting models have demonstrated tremendous potential due to their outstanding performance in long-term time predictions. However, Transformer-based models for multi-variate time series forecasting often come with high time complexity and computational costs. Therefore, we propose a low time complexity model called Fourier U-shaped Network (F-UNet) for multi-variate time series forecasting, which is non-Transformer based. Specifically, F-UNet is composed of low time complexity neural network components, such as Fourier neural operator and feed-forward neural network, arranged in a Ushaped architecture. F-UNet conducts channel and temporal modeling separately for the multi-variate time series. The UNet constructed based on Fourier neural operators is employed to achieve channel interactions, while linear layers are used to realize temporal interactions. Experimental results on several realworld datasets demonstrate that F-UNet outperforms existing Transformer-based models with higher efficiency in multi-variate time series forecasting.

Keywords: Data Mining
Abstract: Machine learning algorithms to aid decision-making processes are increasingly common in several areas., e.g. mobile phones, internet, sensor applications, etc. When fully-trained, algorithms tend to perform better, but the availability of data labels shortly after testing without human intervention is a challenging task in many areas, especially in data stream learning with concept drifts, where data is generated very fast, in real-time, with the possibility of changes in the data distribution. Concept drifts have been addressed in different ways, but using drift detectors with base classifiers in semi-supervised learning is not so common. This article shows how to use state-of-the-art supervised detectors in semi-supervised learning problems, and it also includes an extension to the MOA framework. The Experiments designed to test our proposal used Hoeffding Tree (HT) as base classifier, combined with eight drift detectors and a total of 62 artificial and five real-world datasets, configured with 15% and 30% labeled instances. The results indicate that drift detectors designed for supervised learning can also be effectively used in semi-supervised environments. This finding could lead to a change of paradigm for future research, since supervised drift detectors have never been considered as a viable alternative due to the absence of labels shortly after testing in many real-world data streams.

Keywords: Data Mining, Bio-inspired
Abstract: K-nearest neighbor is one of the simplest and most intuitive binary classification methods providing robust results on a wide range of data. However, classification results can be improved by using a decision method that is capable of assigning, if necessary, the minority label from the list of neighbors of a tested instance. In this paper, we propose using a simple game-theoretic model to assign labels based on the neighbors' information to enhance its performance for binary classification.

Keywords: E-health, Pattern Recognition, Signal Processing
Abstract: Dementia is a chronic and irreversible condition characterized by progressive cognitive and functional decline. While the cognitive and functional decline is profoundly disabling, the non-cognitive Neuropsychiatric Symptoms (NPS) negatively impact these patients' quality of life. In this study, we investigate using artificial intelligence with a wide range of features from wearable sensor data collected from patients with dementia (PwD). Our goal is to develop an assistive artificial intelligence approach that detects NPS in PwD in an institutional setting to understand their behaviors and detect episodes of agitation. We present the preliminary results of a real-world study at Ontario Shores Centre for Mental Health Sciences. The results suggest that using sequential feature selection improved results with fewer features after selecting the optimal features from 198 features. We also found that the Extra Trees classification model can classify the non-agitated normal events and agitated events with the best accuracy compared to other algorithms. Our results demonstrate that the personalized model produced better results, with an average of 5-10% higher than all patients combined.

Keywords: E-health, Deep Learning, Pattern Recognition
Abstract: Detection of seizures using smart cameras has potential benefits since it does not require contact with the patient and can be easily deployed. Timely seizure alerts are crucial to prevent potential complications from seizures such as secondary injuries, and to initiate treatment to stop a seizure. In our small, preliminary study, we demonstrated that camera-based patient-specific seizure detection can provide reliable detection of convulsive seizures and in some cases, even outperform the EEG-based seizure detector. We therefore see a need for the development of user-friendly, trainable, smart camera system which can be easily re-trained for each patient by a caregiver or at home by a family member.

Keywords: E-health, Pattern Recognition, Human-Computer Interactions
Abstract: Musculoskeletal disorders (MSDs) are pervasive in the workforce and constitute the single largest category of work-related illness. The root cause for MSDs is complex. However, there is little dispute that MSD morbidity is primarily due to physical and psychosocial risk factors, and these two domains of risk factors share a common upstream determinant. A work organization influences both the physical load patterns and psychosocial features. In this paper, we propose a technology-facilitated intervention program that could lead to an improved safety culture in workplaces. The program is aimed at addressing one of the physical risk factors, ie rest breaks, and a psychosocial risk factor, ie social support. First, a wearable soft orthosis is used to detect the types of physical activities and load patterns, and to derive an intelligent rest break schedule for each type of activity and load patterns. The orthosis would also remind the participant to take a rest break at appropriate times. Second, a mobile app is developed to cultivate a learning community where the participants could seek and provide social support and increase their awareness of occupational safety. We collected some preliminary app usage data and developed a methodology of identifying app usage patterns using both supervised and unsupervised learning. The feasibility of the method is validated using synthesized data derived from the collected data.

Keywords: Decision Making, Agent-Based Modeling, Explainability
Abstract: This study develops a decision support system for localized epidemiological modelling of infectious disease spread. We propose a Bayesian network topology for performing inference supplementary to an epidemiological simulation framework and a cohesive integration of this decision support system with the framework. The Bayesian network topology is structured with data defined as inputs, outputs, or derived features within the simulation framework. All features are motivated by their clinical relevance and utility for administrative policy guidance. Edges in the final network are quantitatively assessed using structural equation modelling to ensure strong causal connections. Various inference scenarios are demonstrated to provide proof of concept for real-world application and validation in future directions. The outcomes of this project contribute to a larger body of work for infectious disease risk mitigation and emergency management in generalized environments.

Keywords: E-health, Deep Learning, Remote Sensing
Abstract: Cardiovascular diseases are the leading cause of death in the world, with arrhythmias being a significant symptom and risk factor. Advancements in technologies such as low-cost and low-power wearable devices, and machine learning techniques for analyzing big volumes of data offer opportunities to address this issue. However, low-cost devices may have limitations, including reduced data quality due to lower sampling rates, bit depth, and the number of leads recorded. These limitations might produce a significant decrease in machine learning models' performance in detecting arrhythmias. This study investigates the impact of data quality reduction on arrhythmia classification using deep neural network models. High-quality ECG data with 12 leads, 500Hz sampling rate, and 32-bits resolution were transformed into low-quality versions with varying leads (from one to six), 100Hz sampling rate, and 8-bits resolution. Training a state-of-the-art deep learning arrhythmia detection model on both high-quality and low-quality datasets revealed a decrease in performance from 95.3% to 93.9% in the worst case, which is concerning given the critical nature of the domain. To mitigate this performance loss, we propose an ensembling method that compensates for 42% of the loss, achieving an accuracy of 94.5% even with the low-quality dataset. The analysis also identifies the leads with the most promising classification performances. These results can aid in making better design decisions when creating cost-effective wearable ECG devices.

Keywords: Signal Processing, E-health
Abstract: The fusion of inertial and visual data is an effective approach to human motion analysis, with applications in areas such as sports or rehabilitation exercise monitoring. Employing wireless, low-cost, external inertial sensors and a built-in camera on mobile devices provides a convenient acquisition system, available for wide range of potential users. In order to take advantage of both data modalities, robust time synchronization is required. We consider consumer-grade devices, for which direct access to internal clocks is not available and only high-level API is provided. At the same time, we aim to avoid event-based synchronization that would require additional user actions. We investigate sources of acquisition errors on mobile devices, and then we propose and evaluate a novel synchronization method for inertial and visual data. Experimental results indicate that the proposed method provides robust synchronization.

Keywords: Particle Swarm Optimization, Electric Vehicle, Smart Grid
Abstract: The unit commitment (UC) problem is the first step in power system optimal scheduling and system planning. However, the UC problem is a mixed integer optimization problem, which usually has the characteristics of high dimension, non-convex and nonlinear. Plug-in electric vehicles (PEVs) integration into the grid can help improve stability and flexibility of the grid. However, Large-scale PEVs charging demand may put pressure on the grid and may lead to grid overloads. Recently, a competitive swarm optimizer (CSO) is proposed to settle optimization problems, which is considerably challenging in evolutionary computation. In this paper, a binary competitive swarm optimizer (BCSO) is proposed to tackle UC problems integration with PEVs. Finally, comparison experiments on economic problems with dimensionality increasing from 10 to 100 units, which confirm the competitive performance of the proposed optimizer.

Keywords: Smart Grid, Randomized Algorithms, Explainability
Abstract: We propose the concept of ”explainergy”, a new way of including explainability in the metaheuristic performance of algorithms solving problems in the energy domain. To this end, we open the discussion around eXplainable Computational Intelligence (XCI), focusing on using metaheuristic optimization for complex energy-related problems. It is well known that computational intelligence applied to optimization cannot guarantee optimality theoretically and also faces issues related to premature convergence, tuning parameters, and variability of the results. These aspects slow the adoption of such methods by energy industry practitioners. Our proposal considers incorporating ideas already applied to the artificial intelligence paradigm, namely those related to eXplainable AI, to motivate current research in this field and provide solutions from metaheuristics with explainability characteristics. Through a case study solving a bidding problem in local electricity markets, we shed light on some ideas that might be advantageous to understanding the metaheuristic performance for energy experts unfamiliar with approximate algorithms. If an XCI framework is successfully developed, it can increase metaheuristic adoption, reliability, and broader success.

Keywords: Randomized Algorithms, Smart Grid, Advanced Optimization
Abstract: In the energy field, the "WCCI(CEC)/GECCO Competition Evolutionary Computation in the Energy Domain: Risk-based Energy Scheduling" is a platform for testing and comparing new evolutionary algorithms (EAs) to address complex energy problems. Nonetheless, the current competition ranking metric is not statistically significant in assessing algorithm performance and only considers the mean fitness value associated with the objective function. Thus, this work undertakes a statistical analysis using the Shapiro-Wilks test, the Wilcoxon pair-wise comparison, and the Kruskal-Wallis technique to comprehensively study algorithm performance based on statistical grounds. The results reveal that, according to the Wilcoxon test, the top three algorithms demonstrate significant superiority over the other algorithms. In contrast, the Kruskal-Wallis test shows that the top four algorithms belong to the same group based on the ranks resulting from the test. This rigorous analysis provides valuable insights into the stochastic performance of algorithms, contributing to a deeper understanding of their capabilities in the context of the competition.

Keywords: Smart Grid, Particle Swarm Optimization, Intelligent Control
Abstract: As the distributed energy resources (DERs) increasingly penetrate the unbalanced distribution network, it becomes challenging to accommodate such penetration technically and economically. Therefore, this paper tackles an optimal allocation of PV systems (locations and sizes) to maximize the penetration while minimizing voltage violation. It is challenging because the problem is a mixed integer nonlinear programming (MINLP) problem with non-linear and non-convex properties. In addition, the network is unbalanced which brings burdens on solving load flows. Computational intelligent methods, particularly evolutionary algorithms (EAs) have proven its efficiency and robustness in large optimization problems and thus, this paper explores two EAs on the problem with the help of a robust unbalanced load flow algorithm. A comparative study is conducted on particle swarm optimization (PSO) and artificial bee colony (ABC) based on IEEE 13 and 37 bus systems. Optimal allocation based on peak hour and day-ahead scenarios are considered. After 30 times run, the test cases have shown that both EAs are successful and yet ABC generally converges to better solution and yet with larger statistical deviations on solutions.

Keywords: Operations Research, Advanced Optimization, Evolvable Systems
Abstract: Improving buildings’ energy efficiency is an essential component in the efforts for reducing the carbon footprint. The design of more accurate machine learning models for forecasting energy use in buildings can help to reach this goal since these models can be integrated as part of the management systems. A variety of machine learning algorithms have been used for different classes of building energy predictions problems. In this paper we investigate two questions related to the use of neural networks for building energy predictions: The benefits of optimized neural network configurations that include the architecture and some hyperparameters, and the impact on the performance of the amount of data available to train the networks. Our results show that combine optimization of architectures and hyperparameters can significantly improve the accuracy of the neural networks in some problems and that the availability of training data should be taken into account when deciding to apply neural networks over other machine learning methods for building energy prediction problems.

Keywords: Fault Detection, Defense and Security
Abstract: Wireless Sensor Networks (WSNs) are rapidly employed in many applications due to highly demanded autonomous systems. These networks are of immense importance due to their ability to collect data from remote and challenging environments, their impact on various sectors like healthcare, agriculture, industry, environment, and their role in enabling smart technologies for a sustainable, secure, and connected future. Nevertheless, these systems can be attacked by adversaries. Usually, the WSNs are designed with lightweight sensor nodes with limited computation and memory resources. Therefore, employing a firewall system on every sensor node is unacceptable. This paper tackled this problem with a very lightweight Graph Neural Network-based model. The conducted experiment performed in this work demonstrates promising attack-type detection by our proposed approach to the WSN-DS dataset. In this article, our proposed method is compared with other the-state-of-the-art works, and we could discover all Blackhole attacks, one of the most common Denial-of-Service attacks.

Keywords: Defense and Security, Multi-Agent System, Agent-Based Modeling
Abstract: Multi-Agent Pathfinding (MAPF) algorithms represent a powerful tool for modeling realistic tactical movements of troops in military simulation systems. Solving MAPF problems while dealing with topographic terrains involves computing the most cost-effective and safe relief routes for agents with movement constraints. To minimize the overall topographic cost of agents' movement and the need to deviate from other stationary agents, this work considers a MAPF approach that respects real-world agents' movement characteristics, such as the agents' orientation, the limits of the agents' turning angles, and the relief inclinations they can safely navigate. To solve conflicts between agents while navigating uneven terrains, the proposed approach explores the attribution of agents' movement priorities related to the need to execute given missions. Most importantly, other agents without planned movement at the current mission situation, as they are stationary on safe and low-cost routes according to the terrain relief, are minimally displaced to nearby locations to give passage to the mission-critical agents. The MAPF algorithm is evaluated on a comprehensive set of test scenarios, with results analyzed using Generalized Linear Regression models. This analysis provides valuable insights into the MAPF algorithm's effectiveness in virtually modeling organized agent movement behaviors for developing simulation-based training and instruction activities.

Keywords: Federated Learning, Cybersecurity
Abstract: Deep learning and federated learning show significant success in cybersecurity for Intrusion Detection Systems (IDS). This paper presents the Federated Self-Supervised Learning (FSSL) framework proposed for IDSs. FSSL combines Self-Supervised Learning (SSL) with federated learning to obtain a global model. SSL works at the client level, where only unlabeled data is available, and thus it enables the learning from these data. This knowledge enhances the training of the target model. Therefore, FSSL follows a federated learning approach, where private data from multiple clients help to create a global model. Each client learns an unsupervised model, which is then transmitted to a server and combined into a single model. The communication between clients and the server aims to improve model performance and convergence. Conducted experiments compare FSSL with a baseline approach using limited data and a deep learning model. FSSL utilizes an autoencoder to learn a representational model on unlabeled data and transfers knowledge by initializing deep learning model weights with the encoder layers. Results show that FSSL significantly improves the F1-Score of detection systems across three well-known datasets (NSL-KDD, TonIoT, and BotIoT). Moreover, the proposed model demonstrated a noteworthy capability to detect previously unidentified attacks when compared to the baseline.

Keywords: Cybersecurity, Image Processing, Deep Learning
Abstract: We explore gender bias in the presence of facial masks in automated face recognition systems using various deep learning algorithms in this research study. The paper focuses on an experimental study using an imbalanced image database with a smaller percentage of female subjects compared to a larger percentage of male subjects and examines the impact of masked images in evaluating gender bias. The conducted experiments aim to understand how different algorithms perform in mitigating gender bias in the presence of face masks and highlight the significance of gender distribution within datasets in identifying and mitigating bias. We present the methodology used to conduct the experiments and elaborate the results obtained from male only, female only, and mixed-gender datasets. Overall, this research sheds light on the complexities of gender bias in masked versus unmasked face recognition technology and its implications for real-world applications.

Keywords: Deep Learning, Signal Processing
Abstract: In this work we present a novel approach for generating cardiovascular data using a modified WaveNet architecture. This can enable further research in areas where data is scarce and hard to obtain. By generating additional time series data in a set of animal tests performance of existing models could be improved and more difficult approaches, that require substantial amounts of data, attempted. We validate our approach on a classification task and compare it to similar methods of data augmentation.

Keywords: Multi-Agent System, Agent-Based Modeling
Abstract: In Open Multi-Agent Systems (OMAS), heterogeneous agents in varying environments or models can transition from one system to another, retaining their attributes and knowledge. This migration process results in an augmented development complexity compared to conventional Multi-Agent Systems. Additionally, the intricacy of this transition arises from uncertainties and dynamic behaviors associated with the agent's changes, necessitating the formulation of techniques to analyze this complexity and comprehend the system's overall behavior. To address these challenges, we employed Docker, which enables a flexible architecture that accommodates different programming languages and frameworks for the agents. This paper introduces a Docker-based architecture that aids OMAS development, facilitating agent migration between various models operating in heterogeneous hardware and software setups. To validate the proposed approach, we conducted simulations using NetLogo's Open Sugarscape 2 Constant Growback and JaCaMo's Gold Miners. These simulations were executed locally, in the cloud, and in a hybrid mode to assess the feasibility of the proposed architecture.

Keywords: Cybersecurity, Deep Learning, Explainability
Abstract: In network security, applying deep learning methods to detect network traffic anomalies has achieved great results with various network traffic representations. A possible representation is the transformation of raw network communication to images to extract valuable information from the unmanageable amount of network traffic by applying representation learning. However, since deep learning models can result in black boxes for users, it is interesting to understand what valuable information is learned from network communication converted into images. This paper elaborates on that question using explainable artificial intelligence (XAI) methods to identify network packets that most influence the prediction and verify that packets in a malware communication containing malicious payloads have high influence on the prediction. We inspect the Grad-CAM and visualize the Integrated Gradients of Xception and VGG-19 model and investigate the attention heat maps of our Vision Transformer (ViT) model. In addition, we present a novel transformation of sessions to a new image representation to expand the informativeness of network communication. For multiclass classification, our best model Xception achieves an accuracy of 97.95%, whereas, for binary classification, Xception and VGG-19 achieve well above 99.50%. Our ViT model achieves a significantly lower performance with 95.86% for multiclass and 99.36% for binary classification. In particular, computing centers could benefit by examining their inbound and outbound traffic to detect malicious behaviors ahead of time.

Keywords: Cybersecurity, Dimension Reduction, Internet of Things
Abstract: Protecting Internet of Things (IoT) networks from threats is becoming increasingly important as these devices continue to grow in adoption. Modern and unseen attacks that require the analysis of more complex network traffic data for effective identification and mitigation are becoming more prevalent. Traditional machine learning approaches in current intrusion detection systems (IDS) struggle with these volumes of data, prompting exploration into the feature selection space. One class of such feature selection methods is evolutionary algorithms, in which systems mimicking real-life evolution optimize solutions for some problem. In this paper, we propose bag-of-little-bootstraps genetic algorithm feature selection (BLB-GAFS), a novel variant of the genetic algorithm feature selection method that maintains a global search of the solution space while reducing computational cost. This is accomplished with the bag-of-little-bootstraps method for approximating classifier performance. We test the BLB-GAFS technique on three modern intrusion datasets—CCD-INID-V1, detection_of_IoT_botnet_attacks_N_BaIoT, and CIRA-CIC-DoHBrw-2020—that represent updated network patterns and are highly dimensional. We found that the BLB-GAFS method matches or outperforms embedded feature selection methods on the same datasets. Furthermore, the feature sets selected by BLB-GAFS result in significantly improved multiclass precision, recall, and F1-score. Traditionally expensive wrapper feature selection methods like the genetic algorithm can be used on larger datasets through BLB-GAFS, opening the door to other applications with highly dimensional data.

Keywords: Cybersecurity, Defense and Security
Abstract: Vicious assaults, malware, and various ransomware pose a cybersecurity threat, causing considerable damage to computer structures, servers, and mobile and web apps across various industries and businesses. These safety concerns are important and must be addressed immediately. Ransomware detection and classification are critical for guaranteeing rapid reaction and prevention. This study uses the XGBoost classifier and Random Forest (RF) algorithms to detect and classify ransomware attacks. This approach involves analyzing the behaviour of ransomware and extracting relevant features that can help distinguish between different ransomware families. The models are evaluated on a dataset of ransomware attacks and demonstrate their effectiveness in accurately detecting and classifying ransomware. The results show that the XGBoost classifier, Random Forest Classifiers, can effectively detect and classify different ransomware attacks with high accuracy, thereby providing a valuable tool for enhancing cybersecurity.

Keywords: Cybersecurity, Deep Learning
Abstract: Eyeglasses are not only used to protect our vision and prevent dust from getting into our eyes. Additionally, glass that fits properly can give a person an elegant appearance. However, people often find it difficult to choose eyeglasses that fit their face shape; to address this issue, we have proposed a novel architecture in this paper. In order to do this, we created a pipeline that can recommend eyeglasses based on the form of the eyes using multiple transfer learning architecture to predict the face shape from a given image. We utilized InceptionV4 [17], InceptionV3 [18], Vit Small [12], DenseNet121 [10], ResNet50 [9], and VGG16 [16] to predict the facial shape from the image and achieve a test accuracy of 75%. We used 5500 photos with five different face shapes (Heart, Oblong, Oval, Round, Square) for this experiment, and two distinct datasets were gathered from Kaggle [2] and GitHub [1]. By simply uploading the photograph to our recommendation system, our proposed solution can assist users in selecting the appropriate eyewear.

Keywords: Cybersecurity, Defense and Security, Explainability
Abstract: Transformative developments have been brought in across several industries. Digital twin technologies are one of them. This revolutionary innovation has enhanced efficiency, optimized production, and elevated product design to new heights. Nevertheless, as industries embrace the potential of digital twins, cybersecurity concerns come to the forefront due to the convergence of physical and virtual realms. By addressing cybersecurity challenges effectively, industries can fully capitalize on the transformative capabilities of digital twin technology, driving competitiveness and resilience in the face of evolving digital landscapes. Our research explores the various industrial applications of digital twin technology. It also highlights the urgent need for strong cybersecurity measures. Secure data transmission, access control, encryption, and threat detection become crucial elements that must be ensured for digital twin systems in the industrial sector. Our study fills cybersecurity gaps in digital twin applications, offering actionable information for strong security.

Keywords: Cybersecurity, Deep Learning, Smart Grid
Abstract: An increasing number of Intrusion Detection Systems (IDSs) rely on Artificial Intelligence (AI), specifically Machine Learning (ML) algorithms, to distinguish between benign and malicious data and detect cyber attacks. However, using ML algorithms exposes IDSs to Adversarial Machine Learning (AML) attacks during the training and test phase. These AML attacks aim to deceive ML algorithms by misclassifying data, posing significant disruptions to the system and its users. Two critical categories of AML attacks are White-box and Black-box attacks, with Black-box attacks being more practical and representative of real-world scenarios. This paper investigates the impact of adversarial examples on supervised ML models in IDSs and proposes an ensemble learning-based detection approach. The study uses a power system dataset and employs Random Forest, AdaBoost, and Decision Tree classifiers to achieve this. During the test phase, adversarial examples are generated using the decision boundary and HopSkipJump attacks, two types of Black-box decision-based attacks. The research applies a deep neural network to the dataset containing the generated adversarial examples to detect these AML attacks, achieving an accuracy of 98% to 99%.

Keywords: Robotics, Reinforcement Learning
Abstract: In this paper, the robot PID control is compensated by the reinforcement learning. The controller adjustment is proposed by the stability analysis. The reinforcement learning can compensate the dynamics of the robot. This method avoids the problems due to big integral gain of classical PID control. The experimental results show the effectiveness of the proposed controller.

Keywords: Robotics, Particle Swarm Optimization
Abstract: In order to improve the task execution capability of home service robot, and to cope with the problem that purely physical robot platforms cannot sense the environment and make decisions online, a method for building digital twin system for home service robot based on motion simulation is proposed. A reliable mapping of the home service robot and its working environment from physical space to digital space is achieved in three dimensions: geometric, physical and functional. In this system, a digital space-oriented URDF file parser is designed and implemented for the automatic construction of the robot geometric model. Next, the physical model is constructed from the kinematic equations of the robot and an improved particle swarm optimization algorithm is proposed for the inverse kinematic solution. In addition, to adapt to the home environment, functional attributes are used to describe household objects, thus improving the semantic description of the digital space for the real home environment. Finally, through geometric model consistency verification, physical model validity verification and virtual-reality consistency verification, it shows that the digital twin system designed in this paper can construct the robot geometric model accurately and completely, complete the operation of household objects successfully, and the digital twin system is effective and practical.

Keywords: Robotics, Deep Learning
Abstract: Deep Learning (DL) has brought significant advances to robotics vision tasks. However, most existing DL methods have a major shortcoming - they rely on a static inference paradigm inherent in traditional computer vision pipelines. On the other hand, recent studies have found that active perception improves the perception abilities of various models by going beyond these static paradigms. Despite the significant potential of active perception, it poses several challenges, primarily involving significant changes in training pipelines for deep learning models. To overcome these limitations, in this work, we propose a generic supervised active perception pipeline for object detection that can be trained using existing off-the-shelf object detectors, while also leveraging advances in simulation environments. To this end, the proposed method employs an additional neural network architecture that estimates better viewpoints in cases where the object detector confidence is insufficient. The proposed method was evaluated on synthetic datasets - constructed within the Webots robotics simulator -, showcasing its effectiveness in two object detection cases.

Keywords: Decision Making, Human-Like Intelligence, Data Mining
Abstract: Service robots are expected to autonomously perform a wide range of service tasks to satisfy users' needs, but are limited in practice by their weak decision-making capabilities. This work introduces a Knowledge Acquisition Framework (KAFS) to help robots make autonomous decisions through this knowledge. This framework is divided into two parts: service knowledge acquisition and scene knowledge construction and uses a variety of intelligent methods to easily and accurately acquire a large amount of service and scene knowledge. We demonstrate the knowledge acquired by KAFS and validate the effectiveness of KAFS on robot service tasks.

Keywords: Decision Making, Operations Research, Autonomous Systems
Abstract: This paper clarifies the requirements of a hybrid team including both humans and robots, then analyzes and confirms that the Role-Based Collaboration (RBC) methodology and the Environments - Classes, Agents, Roles, Groups, and Objects (E-CARGO) model can meet the requirement and assist in establishing such teams. Following this assessment, this paper proposes to use E-CARGO/RBC in building human/robot teams. Simulations and experiments are used to verify the proposed method.

Keywords: Automated Algorithm, Model-Based, Transportation and Vehicle Systems

Keywords: Bio-inspired, Particle Swarm Optimization, Model-Based
Abstract: Many-task optimization problems (MaTOP) involve more than three tasks, which can be solved simultaneously via knowledge transfer by utilizing complementary information of different tasks. Due to the biases between tasks, relevant tasks are usually selected for knowledge transfer to avoid negative effects. There are two challenging issues, i.e., source task selection and inter-task knowledge transfer. To address these issues, this paper proposes an adaptive geodesic flow kernel transfer method (AGFKTM) for MaTOP. In AGFKTM, multiple source tasks are selected based on both the similarity between tasks and the performance of tasks. In this way, similar and well-performed tasks are selected with a high priority. In addition, an adaptive geodesic flow kernel is constructed to implement knowledge transfer, in which the adopted subspaces along the geodesic flow path are adaptively controlled. Particularly, the transferred solutions are used to generate new ones using mutation operators. Integrating the AGFKTM into differential evolution, a new algorithm named AGFKT-DE is put forward. Experimental results on GECCO20MaTOP benchmark show that the new algorithm outperforms state-of-the-art algorithms.

Keywords: Swarm Intelligence
Abstract: The Kriging surrogate (KS) has been widely used in surrogate-assisted multiobjective evolutionary algorithms (SAMOEAs) for solving expensive multiobjective optimization problems (EMOPs). Typically, when tackling an M-objective EMOP, a KS consists of M Kriging models, in which each model is used to approximate one objective function to replace the expensive fitness evaluation. Since such a KS is only efficient in solving low-dimensional EMOPs, the dimension reduction method has been adopted to construct the reduction surrogate (RS) to reduce training costs. However, both KS and RS can only approximate the mapping from variables to different objectives (i.e., objective function) but ignore the potential relationship between objectives. For practical applications, it is necessary to take into account the mapping between different objectives for the reliability of the surrogate. Therefore, we for the first time propose the concept of the conjugate surrogate (CS) and construct a simple CS to realize the approximated mapping from objectives to objectives. Different from KS or RS, all models in CS are conjugate symbiosis. In collaboration with RS, CS can not only benefit the light training cost, but also improve the convergence speed. Compared with five state-of-the-art SAMOEAs, the CS-assisted algorithm shows its effectiveness and competitiveness in solving EMOPs.

Keywords: Particle Swarm Optimization, Evolving Learning, Swarm Intelligence
Abstract: Multi-objective optimization problems with various attributes are studied for two decades. Sparsity, as one of them, sparked many researchers. However, they usually focused on sparse large-scale bi-objective optimization. The result is unsatisfying when applying their algorithms to optimization problems with more than three objectives. To solve this issue, this paper selects a classical algorithm for large-scale sparse multi-objective optimization problems and proposes the reference points and adaptive crossover and mutation strategies to the original algorithm, adapting it to the sparse many-objective optimization problem. After a series of experiments, the algorithm with this modification mostly dominates other state-of-the-art multi-objective optimization algorithms. Although several best performance metrics are obtained from other competitors, the highest sparsity on the Pareto optimal solution is still completed by the proposed algorithm.

Keywords: Model-Based, Randomized Algorithms
Abstract: Initialization plays a crucial role in surrogate-based multiobjective evolutionary algorithms (MOEAs) when tackling computationally expensive multiobjective optimization problems. During the initialization process, solutions are generated to train surrogate models. Consequently, the accuracy of these surrogate models depends on the quality of the initial solutions, which in turn directly impacts the performance of surrogate-based MOEAs. Despite the widespread use of Latin hypercube sampling as an initialization method in surrogate-based MOEAs, there is a lack of comprehensive research examining the effectiveness of different initialization methods. Additionally, the impact of the number of initial solutions on the performance of surrogate-based MOEAs remains largely unexplored. This paper aims to bridge these research gaps by comparing the usefulness of two commonly employed initialization methods (i.e., random sampling and Latin hypercube sampling) in surrogate-based MOEAs. Furthermore, it investigates how varying the number of initial solutions influences the performance of surrogate-based MOEAs.

Keywords: Advanced Optimization, Model-Based, Data Mining
Abstract: Some real-world engineering problems are offline data-driven mixed-variable optimization problems, which involve optimizing both continuous and discrete variables using only historical experimental data. The main challenges are handling mixed variables and utilizing surrogate models effectively. We propose a novel algorithm that uses a step-wise strategy to optimize the discrete and continuous variables in two stages. In the first stage, we use different radial basis function networks models as surrogates and a voting method to select a promising subspace of discrete variable values. In the second stage, we fix the discrete variable values and use a selective ensemble strategy to optimize the continuous variables. We test our algorithm on 30 test problems and compare it with two representative algorithms. The results show that our algorithm is superior and more stable on most problems, especially on complex multimodal problems. Our algorithm is an effective and flexible framework for handling mixed variables and improving search efficiency and quality.

Keywords: Ambient Intelligence, Decision Making
Abstract: Gaussian processes (GPs) serve as powerful surrogate models in optimisation by providing a flexible data-driven framework for representing complex fitness landscapes. We provide an analysis of realisations drawn from GP models of fitness landscapes—which represent alternative coherent fits to the data—and use a network-based approach to investigate their induced landscape consistency. We consider the variation of constructed local optima networks (LONs: which provide a condensed representation of landscapes), analyse the fitness landscapes of GP realisations, and delve into the uncertainty associated with graph metrics of LONs. Our findings contribute to the understanding and practical application of GPs in optimisation and landscape analysis. Particularly that landscape consistency between GP realisations can vary considerably depending on the model fit and underlying landscape complexity of the optimisation problem.

Keywords: Bio-inspired, Decision Making, Particle Swarm Optimization
Abstract: Dynamic time-linkage optimization problems (DTPs) bring challenges to existing evolutionary algorithms due to the influence of a current decision in the future. Existing methods usually model the rewards of a current decision in the future for prediction. However, these methods often present low prediction accuracy due to the lack of sufficient training data. In addition, they often require a long computational time. To address these issues, the problem of predicting rewards is converted into a simpler binary classification problem, which evaluates whether a current solution can bring positive or negative influence in the future. This paper proposes a clustering-based support vector classifier for solution evaluation. In the proposed method, the density of the time-linkage property is detected first. Historical data are divided using k-means clustering so as to train a support vector classifier for solution evaluation. Good solutions are selected to generate a final decision solution using a crossover operator. Integrating the clustering-based support vector classifier into particle swarm optimization, a new method named CSVC-PSO is put forward. Multiple instances are constructed using a recent DTP test suite with different types of time-linkage patterns and density. Experimental results demonstrate that the proposed CSVC-PSO outperforms state-of-the-art algorithms on most instances using a shorter time.

Keywords: Deep Learning, Particle Swarm Optimization, Remote Sensing
Abstract: The paper presents a novel parameter selection-based image parsing framework that explores additional contextual information to produce final labels. The notable novelties include the optimization of parameters, and the computation of contextual information. The paper demonstrates the improved pixel accuracy, mean pixel accuracy, mean intersection over union and f1-scores using the proposed image parsing architecture. The architecture achieves 84%-pixel accuracy, 77% mean pixel accuracy, mIoU 61% and F1-score of 73% on WHDLD dataset. In comparison to the state-of-the-art techniques the proposed approach achieves better scores. The incorporation of optimization algorithm and the additional context information improves the segmentation accuracies. In our future research, the aim will be to investigate the proposed architecture on a number of real-world image parsing datasets.

Keywords: Advanced Optimization
Abstract: Differential Evolution (DE) is widely recognized as an effective, robust, and gradient-free global optimization algorithm. However, the DE algorithm’s search strategy has certain limitations that present opportunities for further improvement. Opposition-based Learning (OBL) as one of the efficient computational concepts provides the optimizer with the capability of exploring the search space in opposite directions. This research paper introduces a novel crossover scheme for the DE algorithm based on OBL concept. Unlike existing approaches in the literature, which primarily focus on utilization of OBL in population level, proposed scheme takes the advantage of OBL in operation level. In proposed scheme, the crossover operator generates two trial vectors in opposite directions, enhancing the exploration capability of the search strategy and taking a cautious approach by regularly examining the opposite directions during crossover. To evaluate the effectiveness of the proposed method, a series of experiments are conducted using the CEC-2017 benchmark functions with two different numbers of dimensions: 30 and 50. The results demonstrate a significant improvement in performance of the DE algorithm through the proposed method.

Keywords: Model-Based, Advanced Optimization, Decision Making
Abstract: Computationally expensive multiobjective optimization problems (EMOPs) that require significant computational resources are commonly encountered in real-world applications. To address the challenges associated with such problems, using computationally inexpensive surrogate models to approximate objectives has emerged as an effective approach to handle EMOPs. However, the current collaboration between evolutionary algorithms (EAs) and surrogate models is limited, relying on static regression or classification methods that do not fully capture the dynamic evolution process of EAs. This study aims to advance the integration of surrogate-assisted multi-objective optimization by incorporating time-series prediction models. The target is to track the evolutionary trajectory of an EA and enhance its search capability. Specifically, long short-term memory (LSTM) networks are embedded into an EA for surrogate-assisted optimization (SAO). The role of LSTM networks in SAO is thoroughly investigated through ablation studies. Experimental results on six EMOPs demonstrate the potential of using LSTM networks in SAO. The results are compared with those obtained from four representative surrogate-assisted EAs, providing insights into the effectiveness of LSTM-based approaches in addressing EMOPs.

Keywords: Explainability, Deep Learning, Randomized Algorithms
Abstract: Currently, lack of interpretability (or explainability) is one of the major drawbacks for artificial intelligence (AI) models. When we intend to build a physical artificial intelligence (PAI) systems, the model interpretability (MI) becomes a crucial problem. To tackle MI problem, we give the explanation of restricted Boltzmann machines (RBM) from optics theory perspectives in this work. Furthermore, we present a discussion about how to implement optical learning neural network with our developed Optics Theory and Design Methods -- OTDMs. With OTDMs, we can better understand the principle behind the good performance of deep neural networks. OTDMs not only give us an alternative explanation of RBM with optics theories, but also provide the guidance on designing a reliable PAI system also. Consequently, OTDMs pave the road to PAI systems, and make it to become possible for realizing all-optical learning neural network.

Keywords: Swarm Intelligence, Graph Neural Networks, Bio-inspired
Abstract: The robot task allocation is a crucial problem in logistics and distribution where robots are required to perform an array of tasks, with differing locations and numbers. As a result, optimally allocating tasks to available robots to minimize associated costs has become a challenging but essential optimization problem. This paper presents a Graph Convolutional Network (GCN) based Ant Colony Optimization (ACO) denoted as GCN-ACO, to solve the robot task allocation problems, which are formulated as the Travelling Salesman Problems (TSP) and Vehicle Routing Problems (VRP). The GCN-ACO algorithm comprises two stages. In the first stage, a GCN model is trained to predict a heatmap, which represents the probability of each edge belonging to the optimal route within the graph. In the second stage, we integrate the predicted heatmap into ACO to guide the ant colony to select edges with greater potential during the search process.We evaluate our approach's performance through testing on standard TSP and VRP datasets. The experimental results demonstrate that our proposed method has a faster convergence rate and a higher quality of solutions compared to the baseline approaches.

Keywords: Decision Making, Data Mining
Abstract: We study Off-Policy Evaluation (OPE) in contextual bandit settings with large action spaces. The benchmark estimators suffer from severe bias and variance tradeoffs. Parametric approaches suffer from bias due to difficulty specifying the correct model, whereas ones with importance weight suffer from variance. To overcome these limitations, Marginalized Inverse Propensity Scoring (MIPS) was proposed to mitigate the estimator’s variance via embeddings of an action. To make the estimator more accurate, we propose the doubly robust estimator of MIPS called the Marginalized Doubly Robust (MDR) estimator. Theoretical analysis shows that the proposed estimator is unbiased under weaker assumptions than MIPS while maintaining variance reduction against IPS, which was the main advantage of MIPS. The empirical experiment verifies the supremacy of MDR against existing estimators.

Keywords: Swarm Intelligence, Data Mining
Abstract: Nowadays, recommendation systems (RSs) have been widely used in many real-world applications. However, traditional recommendation techniques mainly aim at improving recommendation accuracy, while other metrics to measure the performance of the RSs are not considered. In this paper, a multiobjective recommendation model that considers different metrics, including accuracy, diversity, and novelty of recommendations is established. Compared with recommendation models that only consider accuracy, this model can recommend more different items with higher diversity and more fresh items with higher novelty to enhance the long-term performance of RSs. Moreover, to efficiently solve this multiobjective recommendation model, a multi-population genetic algorithm (MPGA), which follows the multiple populations for multiple objectives (MPMO) framework, is proposed. As far as we know, it is the first time that the advanced MPMO framework is used in RSs. We conduct comparison experiments on three real-world datasets with three state-of-the-art multiobjective recommendation algorithms and two traditional multiobjective evolutionary algorithms. The experimental results indicate that the performance of MPGA is better than all the compared methods.

Keywords: Big Data, Automated Algorithm
Abstract: Human gait prediction is an important task in predictive exoskeleton control. However, if static models are used to facilitate this task, two problems arise. First, the models cannot adapt to new environments and terrains during deployment, and second, the models cannot be personalized to any given end user without costly involvement of a human expert. Incremental models can alleviate these shortcomings, but they usually are prone to catastrophic forgetting, which can be dangerous during live deployment. In this work, we introduce an incremental model, that can learn human gait from scratch without outside interference, but does not fall prey to catastrophic forgetting. We test and evaluate our model on a real world gait database and show, that it delivers competitive results with regard to other standard approaches.

Keywords: Data Mining, Explainability, Deep Learning
Abstract: In numerous cities, population expansion and technological advancements necessitate proactive modernization and integration of technology. However, the existing bureaucratic structure often hinders local officials' efforts to effectively address and monitor residents' needs and enhance the city accordingly. Understanding what people find important and useful can be inferred from their posts on social media. Twitter, as one of the most popular social media platforms, provides us with valuable data that, with the right tools and analysis, can provide insights into the performance of urban services and residents' perception of them. In this study, we used the city of Calgary as an exemplar to gather tweets and analyze topics relating to city development, urban planning, and minorities. Natural language processing (NLP) techniques were used and developed to preprocess stored tweets, classify the emotions, and identify the topics present in the dataset to eventually provide a set of topics with the prevalent emotion in that topic. We utilized a variety of methods to analyze the collected data. BERTopic for topic modeling and few-shot learning using Setfit for emotion analysis outperformed the others. Hence, we identify issues related to city development, senior citizens, taxes, and unemployment using these methods, and we demonstrate how delving into these analyses can improve urban planning.

Keywords: Data Mining
Abstract: Feature extraction is essential in bioinformatics because it transforms genome sequences into the feature vectors required for data mining activities such as classification and clustering. The data mining activities enable us to classify or cluster the newly sequenced genome to the known families. Nowadays, a variety of feature extraction strategies are available for genome data. Nevertheless, several existing algorithms do not extract context-sensitive key properties, also some approaches extract features, which are unable to distinguish between two non-similar sequences. In addition, the efficacy of existing feature extraction techniques is evaluated on either supervised or unsupervised learning models, but not on both. Thus, an efficient feature extraction technique that extracts significantly relevant features from genome sequences is required. In this paper, a novel feature extraction method is proposed that extracts features based on the length of the sequence, the frequency of nucleotide bases, the modified positional sum of nucleotide bases, the distribution of nucleotide bases, and the entropy of the sequence to generate a 14-dimensional fixed-length numeric vector to describe each genome sequence uniquely. By applying extracted features to both supervised and unsupervised machine learning approaches, the performance of the proposed feature extraction method is assessed. The experimental results show that the proposed strategy for clustering and classifying novel genome sequences into recognized genome classes is highly effective and efficient. The same is proven by comparing the proposed method to the standard state-of-the-art method.

Keywords: Ambient Intelligence, Data Mining, Big Data
Abstract: Annually, approximately 500,000 Merger and Acquisition (M&A) transactions are disclosed globally, each transaction inciting substantial perturbations to the associated companies’ equity prices. The probability of an M&A transaction’s closure, as perceived by the public, inherently influences the stock price of the target company leading up to the proposed date of the deal. Given the recent advancements in the realm of Natural Language Processing (NLP), we propose an empirical investigation into the correlation between digital dialogue surrounding M&A transactions and consequent movements in the stock prices of involved companies. Utilizing transformer-based encoder-only architectures, we fine tune a stance detection model on an extensive dataset, amassed from digital communication platforms, featuring public discourse related to five historical M&A transactions. Ultimately, we achieved 70% accuracy on deal-completion stance detection using the Roberta-base model. We subsequently employ the aggregated the public sentiment towards the completion or termination of a proposed M&A transaction to model stock price movement. Utilizing a multitude of time series based approaches, we achieve a mean absolute error of 2.29 USD for next-day price prediction and 3.40 USD for next-week price prediction. Ultimately, we find an existing but tenuous relationship between online discourse and the price trajectory of target companies, ultimately highlighting the complex social and economic phenomena behind M&A deals.

Keywords: Deep Learning, Pattern Recognition, Ensemble Learning
Abstract: This paper presents an approach for detecting unilateral arm paralysis/weakness using kinematic data. Our method employs temporal convolution networks and recurrent neural networks, guided by knowledge distillation, where we use inertial measurement units attached to the body to capture kinematic information such as acceleration, rotation, and flexion of body joints during an action. This information is then analyzed to recognize body actions and patterns. Our proposed network achieves a high paretic detection accuracy of 97.99%, with an action classification accuracy of 77.69%, through knowledge sharing. Furthermore, by incorporating causal reasoning, we can gain additional insights into the patient's condition, such as their Fugl-Meyer assessment score or impairment level based on the machine learning result. Overall, our approach demonstrates the potential of using kinematic data and machine learning for detecting arm paralysis/weakness. The results suggest that our method could be a useful tool for clinicians and healthcare professionals working with patients with this condition.

Keywords: E-health, Pattern Recognition, Signal Processing
Abstract: This preliminary study uses a fine-tree machine learning algorithm to replicate bruxism biofeedback systems by detecting bruxism episodes using a wearable sensor system. The detection of bruxism grinding was performed among five different resting/sleeping positions--laying on the front, back, left, and right, and sitting up from four participants. A sequence of ten activities (each activity is a combination of sleeping position and grinding or not grinding) was recorded while wearing the wireless sensing system on the front of the chin directly under the mouth. Both time and frequency domain features were extracted from each axis of the wearable sensor system’s accelerometer data sets. They were used to determine the presence of teeth grinding with 98% accuracy, and these features were used and experimented with to optimize the classification accuracy of the system.

Keywords: Dimension Reduction, E-health, Pattern Recognition
Abstract: DNA sequencing is the process of reading individual base pairs from a section of DNA. Genes are the name given to parts of the DNA which encode proteins; for example ion channels are proteins that maintain concentrations of ions within cells. The sequencing of these genes can offer insights into factors such as evolution and disease. During the sequencing process, unknown values 'N' can be substituted in the sequence where the sequencing machine is unable to identify a nucleotide as Adenine (A), Cytosine (C), Thymine (T), or Guanine (G). These gene sequences vary in length; this includes individual genes across the same species. This has led to the use of a process known as k-mer encoding so that a machine learning algorithm can assess these genes without the need for pre-alignment. K-mer encoding works by taking small sections of the sequence and tallying the number of times that such a sequence appears, such as, how many times the k-mer 'ACCT' appears in the overall sequence. The unknown 'N' value presents a problem in k-mer encoding, as this value increases the size of the k-mer feature vector exponentially as the k-mer length increases. In this paper we research the accuracy and computational impact of including, removing, or ignoring this 'N' value for the k-mer lengths 3, 6, and 9 across four Machine Learning algorithms: Random Forest, Multinomial Naive Bayes, Neural Networks, and Linear Support Vector Machine.

Keywords: Biometric Systems
Abstract: In biomechanics, the calculation of inertial parameters for the upper and lower limbs is studied for motion analysis or the design of prostheses or exoskeletons. However, the calculation of inertial parameters for the hand is performed without considering that the geometry of this segment can change depending on the posture. This work presents a geometric method based on the kinematic model to estimate the inertial parameters of the hand segment for different hand postures. The resulting inertia tensor is calculated at the center of mass according to the segment axes' International Society of Biomechanics (ISB) designation. It considers the principal moments of inertia and the products of inertia of the hand segment. To demonstrate the use of this tool, six healthy subjects participated. The anthropometric measurements of their hand were obtained, the inertial parameters were calculated with our proposal, and they were compared with two methods, Dumas and De Leva, using the Euclidean and Frobenius norms for the center of mass and the inertia tensor, respectively. The mean difference and SD between the proposed method for the relaxed hand position against the Dumas method is 0.0049 m (SD 0.002) and 0.00016·(10e−3) kg−m2 (SD 0.00009) and the De Leva method is 0.011 m (SD 0.0013) and 0.00023·(10e−3) kg−m2 (SD 0.00004) for the center of mass and the inertia tensor, respectively. However, our method can be extended to different hand positions. The proposed method can be used in applications such as the analysis of the three-dimensional motion of the upper limb or in the design of biomedical devices such as hand or wrist and forearm exoskeletons.

Keywords: Human-Computer Interactions, Ensemble Learning, Signal Processing
Abstract: The hand has excellent functional, aesthetic and social importance. However, Parkinson's disease, stroke, and other myopathies can cause motor impairments. Patients require a rehabilitation program to follow their progress, and one of the tools used to do that is the electromyographic (EMG) signals. This article proposes using algorithms to characterize and classify EMG signals during force exercises for the wrist and forearm. Eight healthy subjects participated in this study. They performed seven exercises, making five trials for each one. Signal features were analyzed in different time windows using a genetic algorithm and machine learning techniques to select the window that maximizes the classification. Combining four electrodes, seven exercises, and 14 algorithms achieved a classification accuracy of 92.41% using the Multilayer Perceptron classifier. The study demonstrates a highly reliable method for classifying forearm and wrist exercises based on EMG signals, useful for exoskeletons or rehabilitation platforms. Future work will focus on implementing EMG signals to enhance motor rehabilitation therapy and provide findings that will help the scientific community investigate the combination of EEG signals for rehabilitation purposes.

Keywords: Bio-inspired, Data Mining, E-health
Abstract: Nowadays, gait data analysis has become an extremely valuable tool that, without much knowledge, provides significant support in various areas, especially in medicine. This type of analysis not only contributes to generating accurate diagnoses but also plays a fundamental role in physical rehabilitation processes.

To harness the potential of this analysis, a multimodal system is being developed with the purpose of enhancing the storage, analysis, and synchronization of multiple modules. The imperative arises from certain constraints within prevailing devices and methods, which stem from their intricate and delicate nature. Therefore, the aim of the study involves creating an implementation that prioritizes flexibility, lightness, and autonomy, all with the ultimate aim of attaining complete self-sufficiency in future advancements.

Keywords: Evolvable Systems, Robotics, Model-Based
Abstract: This study investigates how the introduction of morphological novelty affects the fitness and diversity of a population of modular robots. Novelty is usually measured in behavioral space, while the approach discussed in this paper assesses novelty solely using morphologies. The proposed algorithm is inspired by the histogram of oriented gradients, in combination with elements of principal component analysis, and the Wasserstein distance. The experiments conducted utilize novelty in parent selection, with different configurations. Analyzing the results, the introduction of morphological novelty promotes beneficial effects on fitness and diversity when applied correctly.

Keywords: Robotics, Evolvable Systems, Deep Learning
Abstract: A key challenge for evolving complex physical objects is to design a representation, that is, to devise suitable genotypes and a good mapping from genotypes to phenotypes (the objects to be evolved). This paper outlines a new approach to address this challenge for evolving robot morphologies and presents a proof-of-concept study to assess its feasibility. The key idea is to design genotype-phenotype mappings using variational autoencoders. This idea is implemented and tested for the evolution of modular robots for a locomotion task. The experiments show the practicability of this idea where the representation is not hand-designed, but algorithmically generated. This indicates a great future potential for the evolution of complex objects where there are no straightforward representations to use.

Keywords: Evolvable Systems, Bio-inspired
Abstract: Engineering design has traditionally involved human engineers manually creating and iterating on designs based on their expertise and knowledge. In Bio-inspired Evolutionary Development (EvoDevo), generative algorithms are used to explore a much larger design space that may not have ever been considered by human engineers. However, for complex systems, the designer is often required to start the EvoDevo process with an initial design (seed) which the development process will optimise. The question is: will a good starting seed yield a good set of design solutions for the given problem? This paper considers this question and suggests that sub-optimal seeds can provide, up to certain limits, better design solutions than relatively more optimal seeds. In addition, this paper highlights the importance of designing the appropriate seed for the appropriate problem. In this paper, the problem analysed is the structural performance of a Warren Truss (bridge-like structure) under a single load. The main conclusion of this paper is that up to a limit sub-optimal seeds provide in general better sets of solutions than more optimal seeds. After this limit, the performance of sub-optimal seed starts to degrade as parts of the phenotype landscape become inaccessible.

Keywords: Evolvable Systems, Bio-inspired, Multi-Agent System
Abstract: Evolutionary approaches to engineering design involve generating populations of candidate solutions that compete via a selection process iteratively, to improve measures of performance over many generations. Although the attractive properties of biological evolutionary systems have motivated researchers to investigate emulating them for engineering design, there has been an emphasis on using encodings of the technical artefacts themselves, rather than encoding a complete bio-inspired system which is capable of producing such artefacts. It is the latter approach which is the subject of this contribution: how might a bio-inspired system be designed that self-organises the process of engineering design and manufacture? To make progress in the application of evolutionary processes to problems in engineering design, the evolutionary model must encompass the complexity of systems engineering. A new theory of evolutionary systems engineering is presented, based on von Neumann’s Universal Constructor Architecture (UCA), drawing from more recent understanding of biology and applying the resulting system to the task of engineering design. It demonstrates how individual bioinspired algorithms fit into a coherent whole, and how they can be combined to drive open-endedness in automated design. The resulting system provides a common language for multidisciplinary applications in generative design, whereby industrial systems engineering approaches can be developed using principles from the UCA for the first time.

Keywords: Robotics, Bio-inspired, Evolvable Systems
Abstract: This work investigates how a predator-prey scenario can induce the emergence of Open-Ended Evolution (OEE). We utilize modular robots of fixed morphologies whose controllers are subject to evolution. In both species, robots can send and receive signals and perceive the relative positions of other robots in the environment. Specifically, we introduce a feature we call a tagging system: it modifies how individuals can perceive each other and is expected to increase behavioral complexity. Our results show the emergence of adaptive strategies, demonstrating the viability of inducing OEE through predator-prey dynamics using modular robots. Such emergence, nevertheless, seemed to depend on conditioning reproduction to an explicit behavioral criterion.

Keywords: Evolvable Systems, Robotics, Autonomous Systems
Abstract: Small Flapping-Wing Micro Air Vehicles (FW-MAVs) may be subjected to either or both of manufacturing defects or in-service damage that render their pre-designed controllers less than adequately effective. Even minor damage to wings, for example, can remove the vehicle’s ability to reliably follow waypoint trails even if that same damage does not result in a catastrophic loss of altitude. One solution to this problem is to adapt the core wing motion scripts (wing gaits) in an attempt to use wing motion to compensate for losses of force and torque generation due to in-service damage or manufacturing faults. This approach presents a number of challenges - especially if it is to be deployed in a resource restricted vehicle in an online mode during an actual mission. Prior to this paper, we had presented only anecdotal treatment of fully unrestricted, 3D flight. In this paper, we will definitively establish the utility of EA adaptation of flight control in unrestricted flight in a pendulum-stable FW-MAV. We will, additionally, introduce mutation and crowding modifications that provide demonstrable utility in a manner amenable to implementation on a resource-restricted micro vehicle. The paper will conclude with a discussion of open-issues and the potential application of the reported methods to other problems.

Keywords: Human-Like Intelligence, Explainability, Pattern Recognition
Abstract: As the use of artificial intelligence (AI) models becomes more prevalent in industries such as engineering and manufacturing, it is essential that these models provide transparent reasoning behind their predictions. This paper proposes the AI-Reasoner, which extracts the morphological characteristics of defects (DefChars) from images and utilises decision trees to reason with the DefChar values. Thereafter, the AI-Reasoner exports visualisations (i.e. charts) and textual explanations to provide insights into outputs made by masked-based defect detection and classification models. It also provides effective mitigation strategies to enhance data pre-processing and overall model performance. The AI-Reasoner was tested on explaining the outputs of an IE Mask R-CNN model using a set of 366 images containing defects. The results demonstrated its effectiveness in explaining the IE Mask R-CNN model's predictions. Overall, the proposed AI-Reasoner provides a solution for improving the performance of AI models in industrial applications that require defect analysis.

Keywords: Big Data
Abstract: Data augmentations have been shown to improve predictive performance of machine learning models in many domains. Augmentations are typically used to improve classification performance, but augmentations can distort the intrinsic properties of the original data, thus reducing the utility of a model for real-world applications. Because augmentations directly affect the training data, and thus also affect the machine learning models trained with said data, intelligent selection of augmentations is as critical as the selection of input features and other options in the machine learning pipeline. Such an approach will enable greater transferability of trained models from the research lab to products and services.

This paper presents two metrics to evaluate the potential and realized impact of data augmentations. The first metric, eff-score, assesses the relative efficacy of prospective data augmentations before model training. To observe augmentation effects on the intrinsic properties of the training data, the second metric, nirvana distance, measures the effect of data augmentations beyond overall predictive performance after model training. These metrics are tested with a well known multi-purpose audio data set and augmentations from the domain of environmental sound scene analysis. The relative eff-scores correlate with classification results from predictive models trained on the augmented data sets, and the distance components of the nirvana distance explain results observed but not previously understood from output confusion matrices. These results demonstrate promise for data-driven, efficient selection of data augmentations whilst exposing previously hidden impacts on machine learning models. Furthermore, since eff-score and nirvana distance are domain-independent, these metrics have widespread applicability.

Keywords: Image Processing, Dimension Reduction, Pattern Recognition
Abstract: The growing field of Content-Based Medical Image Retrieval (CBMIR) plays an integral role in the diagnosis and treatment plan of numerous diseases, including cancer. However, the effective representation of gigapixel medical images via a large number of extracted features remains a challenging task, crucially influencing other processes in a digital workflow. In this paper, we propose a novel QR code representation strategy to enhance retrieval performance. Unlike the traditional one-dimensional binary representation methods (i.e., barcodes), this 2D approach captures more intricate and informative patterns from the features by differentiating each pair of features resulting in a 2D binary vector. We delve into three distinct QR code generation strategies, namely, the Thresholding QR, the MinMax QR, and the Hybrid QR, each offering unique strengths. Our experiments on representing whole slide images of the Cancer Genome Atlas (TCGA) dataset reveal that while the Hybrid QR tends to provide balanced performance, there are instances where the other two methods outshine. Even though this approach requires more memory usage, the considerable enhancement in accuracy justifies this trade-off; obviously, in medical applications, accuracy holds the highest priority. Hence, the findings indicate that QR codes can effectively improve the performance of CBMIR systems by not only accelerating the retrieval process but also increasing the accuracy of image retrieval, leading to potentially more accurate diagnoses and treatment planning.

Keywords: Deep Learning, Autonomous Systems, Data Mining
Abstract: Email is one of the most common methods of official and personal communication to exchange information. For the administration department, dealing with hundreds of emails with the same type of inquiries or requests results in a huge operational overhead. In this study, we explore email classification models. An email classification system should understand the topics in the email content for categorizing emails and indicate if an incoming email should be handled by the mailbox owner. Email categorization based on topics is a multi-label classification task. Most existing email categorization models perform binary classification to identify spam, phishing, or malware attacks. We propose a CNN-BiLSTM model for multi-class email classification. Our experiments show that compared to the two other models that we implemented namely CNN (76.19%) and BiLSTM (61.9%) models, the CNN-BiLSTM (83.33%) and Hierarchical CNN-BiLSTM models (85.33%) have much better performance.

Keywords: Data Mining, Pattern Recognition, Decision Making
Abstract: Metagenomic analysis is vital in determining what organisms are present in a microbial sample and why they are present. In this study, we explore the utility of MMseqs2, a bioinformatics pipeline, for taxonomic classification in metagenomics, focusing on 16S rRNA gene sequences. We evaluate the algorithm’s performance in full dataset as well as batch-by-batch incremental processing, and more importantly, we add the capability of semi-supervised classification to this otherwise clustering-only algorithm. Incremental updating is important because it allows seamless integration and processing of new data, whereas semi-supervised classification allows taxonomic identification of previously unknown organisms. We also evaluate the different clustering modes offered by MMseqs2, and compare MMseqs2 to our previously developed semi-supervised incremental algorithm SSI-VSEARCH. We show that MMseqs2’s built-in clusterupdate function works well, and our semi-supervised classification capability adds new functionality to this bioinformatics processing pipeline.

Keywords: Reinforcement Learning, Deep Learning, Image Processing
Abstract: Image captioning applications require prompt and precise caption generation, which can improve the accessibility and understanding capabilities of images. We utilize an actor-critic approach based on deep reinforcement learning and propose a two-fold approach to enhance the performance of the actor-critic approach. First, we propose a novel image-text matching module to compute the reward in image-matching, for the actor-critic model. This module enables more accurate and meaningful evaluations, contributing to improved caption generation. Second, we apply various training scenarios in reinforcement learning, strategically updating both the policy and value networks. The scenarios ensure more effective learning dynamics and lead to enhanced overall performance. To assess the efficiency of our approach, we employ the Microsoft COCO dataset. The experiments demonstrate the superiority of our method in terms of both speed and precision compared to the existing techniques.

Keywords: Cybersecurity, Fuzzy Systems, Defense and Security
Abstract: We present a novel access control middleware for distributed multi-user review. The system uses a fuzzy inference system trained on real world access control rules to evaluate and select reviewers as an extension to a more traditional access control system. The method is intended for high security need specific requests, as a supplement to regular access control methods. In this way, it models a multi-person access system common in mechanical controls like missile launches, bank vault opening, and other high criticality domains. The proposed method improves security by increasing the number of compromised users needed to perform an attack, taking advantage of situational awareness of peer users in a system. We evaluate the proposed system with an example implementation based on a real-world organization, and show that the system can be used to effectively implement a secure resource access control system. Our work contributes to the growing body of research into fuzzy-logic access control, ML in access control, and multi-user authentication systems.

Keywords: Cybersecurity, Internet of Things, Evolving Learning
Abstract: Abstract—Internet of Vehicles (IoV) represents the application of Internet of Things (IoT) within vehicular communication environments. Internet of vehicles refers to a network of interconnected sensors, network layers, and communication systems that enable vehicles to connect with everything (V2X communication). IoV networks face numerous security challenges due to the emergence of modern types of attacks with unusual patterns. Therefore, it is a crucial and demanding task to design intelligent Intrusion Detection Systems (IDSs) for IoV networks. In this paper, we propose an optimized Machine Learning-based IDS to detect attacks in IoV networks. We deploy highly efficient Machine Learning models, Light Gradient Boosting Machine, Extra Trees Classifier, and Extreme Gradient Boosting, to detect attacks in the CICDDoS2019 dataset. We apply the Synthetic Minority Oversampling Technique to resolve the issue of imbalanced data distribution of target class. A Correlation-based Feature Selection is conducted to reduce the computational cost by decreasing the number of input variables. In order to enhance the performance of the attack detection, hyperparameters are optimized using the Bayesian Optimization algorithm. The performance evaluation results show that these ML models perform well. Notably, the Extreme Gradient Boosting classifier outperforms other Machine Learning models, and our proposed solution outperforms existing systems in terms of Accuracy score.

Keywords: Defense and Security, Autonomous Systems, Cybersecurity
Abstract: Artificial intelligence (AI) is not a fancy term anymore, or not limited to only researchers and academia. AI is currently becoming a part and parcel of our daily life, we are using AI/ intelligent systems by knowing or without knowing. Event product manufacturers are also trying to incorporate AI with their products to make it more preferable to consumers and trying to get the full benefit of using AI in their production and control units even for business decisions. Therefore, In our paper, we give a comprehensive survey of recent advances in Computational intelligence in industrial Control Systems and cover many usages of how industrial Control Systems are getting benefits from using Computational intelligence. We covered multiple domains like Manufacturing, Energy Management, Transportation, Food and Beverage Industry, and Pharmaceutical Industry, how these industries are utilizing multiple CI-based control systems like Programmable Logic Controllers, Distributed Control Systems, Supervisory Control and Data Acquisition, Industrial Automation, and Control Systems, Intelligent Electronic Devices and found benefits in their operations and manufacturing which helping them to focus more in innovation and improvement of their products. We believe that this survey shall be valuable to researchers across academia and industry.

Keywords: Cybersecurity, Defense and Security, Fault Detection
Abstract: This paper presents a comprehensive analysis of the security aspects of the InsightFace project (a popular open-source face recognition system) focusing on its susceptibility to three distinct black box attacks: Face Swap, Morphing, and Presentation. Open-source face recognition models are used in commercial applications, thereby motivating our security analysis. Our investigation entails a meticulous evaluation of the susceptibility of the project to false authentication when subjected to the three attacks. We observed from our experiments that InsightFace was not able to differentiate between legitimate images and manipulated images. The principal aim of this research is to draw attention to the security challenges inherent in open-source face recognition systems, often integrated into various public applications.

Keywords: Federated Learning, Ethical AI, Cybersecurity
Abstract: Federated learning (FL) enables collaborative model training while keeping data decentralized. However, system heterogeneity and statistical differences in decentralized data can introduce biases and unfairness. This paper surveys existing bias mitigation techniques in FL across various phases of the training process. We identify sources of bias and present a critical analysis of current fairness-aware FL algorithms, categorizing them as preventive (Pre-processing) or reactive (in-processing and Post-processing) based on when bias mitigation is applied. In addition, this paper reveals open challenges in balancing fairness and efficiency in FL, handling non-independent and identically distributed (non-IID) data, and ensuring privacy. This survey lays out the foundation for developing unbiased and privacy-preserving FL systems without discrimination in the future.

Keywords: Cybersecurity, Electric Vehicle, Dimension Reduction
Abstract: Controller Area Network bus systems within vehicular networks are not equipped with the tools necessary to ward off and protect themselves from modern cyber-security threats. Work has been done on using machine learning methods to detect and report these attacks, but common methods are not robust towards unknown attacks. These methods usually rely on there being a sufficient representation of attack data, which may not be available due to there either not being enough data present to adequately represent its distribution or the distribution itself is too diverse in nature for there to be a sufficient representation of it. With the use of one-class classification methods, this issue can be mitigated as only normal data is required to train a model for the detection of anomalous instances. Research has been done on the efficacy of these methods, most notably One-Class Support Vector Machine and Support Vector Data Description, but many new extensions of these works have been proposed and have yet to be tested for injection attacks in vehicular networks. In this paper, we investigate the performance of various state-of-the-art one-class classification methods for detecting injection attacks on Controller Area Network bus traffic. We investigate the effectiveness of these techniques on attacks launched on Controller Area Network buses from two different vehicles during normal operation and while being attacked. We observe that the Subspace Support Vector Data Description method outperformed all other tested methods with a Gmean of about 85%.

Keywords: Swarm Intelligence, Particle Swarm Optimization, Model-Based
Abstract: This paper has two objectives. Firstly, to introduce a new framework XF-OPT/META for testing and comparing Hyperparameter Optimization (HPO) methods. The framework supports model-free methods, e.g., Random Search (RS), as well as model-based methods, such as Bayesian Optimization (BO), with various surrogate models. Due to the generalized and modular structure of the XF-OPT/META framework, it can be easily extended to other optimization methods for different optimization problems. The second objective is to empirically compare the performance of various HPO methods for population-based metaheuristics. For that the XF-OPT/META framework is used to apply HPO methods to the Hierarchical Simple Probabilistic Population-Based Optimization (H-SPPBO) metaheuristic for the Dynamic Traveling Salesperson Problem (DTSP) and to calculate high-performing parameter values for H-SPPBO. Promising results are obtained using the parameter values found by BO. In particular, a parameter set obtained with Gradient-Boosted Regression Trees (GBRT) outperforms a reference parameter set for H-SPPBO from an existing study.

Keywords: Particle Swarm Optimization, Swarm Intelligence, Bio-inspired
Abstract: Particle swarm optimization (PSO) is a swarm intelligence algorithm that finds candidate solutions by iteratively updating the positions of particles in a swarm. PSO performance depends on the use of a suitable control parameter (CP) configuration, which governs the trade-off between exploration and exploitation in the swarm. Various methods of adapting or tuning CPs exist, but many result in exploding particle velocities and an unstable search process. Poli’s stability condition ensures convergence in the mathematical limit, and is often used to inform CP configuration. However, this study shows that since it does not place any practical convergence constraints, it cannot be used to guarantee a stable search process. Velocity explosion occurs nonetheless and can lead to floating-point overflow and numerical instability. The investigation into various CP configurations across diverse functions and measurements of particle velocities provides empirical evidence of velocity explosion, and cautions against the assumption that enforcing Poli’s criterion guarantees stability. The findings underline the need for comprehensive understanding of CP tuning and stability conditions in PSO, as well as the crucial role of empirical evidence in evaluating the real-world performance of swarm intelligence algorithms.

Keywords: Swarm Intelligence, Human-Computer Interactions, Human-Like Intelligence
Abstract: The paper proposes a framework of systems for creating intelligent behaviors of imaginary creatures for humans, which is built upon a framework of swarm intelligence optimization algorithms. The paper assumes and models an imaginary slime mold, which is an amoeboid unicellular creature, as an imaginary creature in the framework concretely. In addition, the paper conducts basic evaluations of the framework. Various swarm intelligence optimization algorithms have been proposed so far, but there is a common feature among them. That is that a set of search points in a search space, which are called individuals, move around in the space according to algorithm specific rules using fitness values of the individuals, and the differences among algorithms are in algorithm specific rules. Therefore, under the use of one swarm intelligence optimization algorithm, that is, one particular set of rules, if a fitness function is varied, behaviors of individuals are also varied. Based on this fact, the proposed framework optimizes a parametric fitness function for individuals to behave intelligently for a human. In the basic evaluation of the concrete system, a fitness function of computer program which returns a fitness value calculated with a distribution of individuals is used instead of a human, and it is demonstrated that optimization of a parametric fitness function indeed yields desired behaviors of individuals.

Keywords: Swarm Intelligence, Advanced Optimization, Explainability
Abstract: We propose a new swarm intelligence numerical optimization algorithm that represents individuals as dynamic graphs in the Euclidean search space. We call it Graph Building Optimization Algorithm or GBO. The unique point of GBO is that an individual is represented by a dynamic graph whose nodes have coordinates (search points) in the Euclidean search space. Due to this unique point, we can draw a GBO's search process as a generation-transition of a feature of a graph. It is expected that we can obtain better understandings on a given problem by comparing the generation-transition for the given problem to the baseline for the simplest unimodal problem. We assume the maximum node degree in the best individual as the feature and the generation-transition of the feature for F1 in the CEC'13 test problems as the baseline. We demonstrate that we can guess the characteristics of other 27 problems in the CEC'13 test problems by comparing their generation-transitions to the baseline. In addition, we evaluate GBO using the same problems and show that GBO is capable of finding good solutions for various problems.

Keywords: Swarm Intelligence, Dimension Reduction, Bio-inspired
Abstract: This study proposes a multimodal approach to enhance the Improved version of Binary Fish School Search (IBFSS) algorithm by incorporating aspects of the Weight Based Fish School Search (wFSS) algorithm to address the attribute selection problem. The proposed model, named Weight Binary Fish School Search (wBFSS), was evaluated on three benchmark datasets, consistently delivering the best solutions in most of the runs. Additionally, two variations of the new wBFSS model were tested to understand the impact on the algorithm's performance by adding a fitness function evaluation before executing the Collective Instinctive Movement.

Keywords: Swarm Intelligence, Multi-Agent System, Autonomous Systems
Abstract: Coverage holes are a key problem in wireless sensor networks. Methods that use relative localization techniques to restore the service, or heal the holes, rely on accurate range and bearing measurements. However, high-precision range and bearing sensors are too heavy, expensive, and range-limited for the agents tasked with healing. To overcome these limitations, we propose a novel impressionist algorithm, inspired by a recent swarm-based approach, that works with extremely coarse range and bearing information and at low perception frequency, to detect and heal the holes. In the proposed approach, a swarm of agents uses quantized information to navigate a potential field, generated by network nodes, to reach the nearest hole. The swarm adopts a greedy deployment behavior, preventing concurrent placement in close-by locations. After deployment, agents use their coarse perception to update the potential field, leading the rest of the swarm to unhealed area. Simulation results demonstrate that our algorithm achieves similar or better coverage compared to the state-of-the-art and to a benchmark based on random walk. This is achieved using just three bearing quantization levels and four times lower perception frequency. Overall, our impressionist approach shows faster healing, albeit at the expense of employing slightly more agents.

Keywords: Transportation and Vehicle Systems, Advanced Optimization, Automated Algorithm
Abstract: Automated Guided Vehicles (AGVs) enhance transportation efficiency in different domains such as warehouses, factories, and container ports. Much research has been done into optimal scheduling and routing of multiple AGVs to improve the overall efficiency of the systems. However, more research efforts are required when addressing more complex real-life systems where the mobility of AGVs is highly constrained due to special geometric shapes and dimensions. Focusing on a real-world hospital AGV routing problem, this paper tackles the additional complexity arising from space capacity constraints long narrow corridors and lifts for cross-floor deliveries. A simulation optimisation approach is introduced to accurately model complex interactions of AGVs under conditions like floor switching, charging, and passing narrow corridors. To tackle the underlining vehicle routing problems with pickup and delivery (VRPPD) which is NP-Hard, this paper presents a simulation-based hyper-heuristic optimization approach to minimize the makespan of all tasks. A surrogate model is integrated to expedite the search process, and several experiments are conducted to properly evaluate the performance of our method. Based on the results, our method exhibits great potential in improving efficiency while maintaining the excellent practicality of AGV routing for complex environments like hospitals.

Keywords: Bio-inspired
Abstract: This paper proposes a quantum representation-based genetic algorithm for solving the job-shop scheduling problem, aiming to minimize the makespan. The job-shop scheduling is a typical scheduling problem that falls under the NP-hard combinatorial optimization problems and has undergone extensive investigation in the literature. Over time, various heuristic and intelligent methods have been developed to tackle this challenging problem. Inspired by the promise of quantum computing, this paper explores using quantum information representation and processing techniques to enhance the performance of conventional genetic algorithms on classical computers to solve the job-shop scheduling problem. The proposed quantum-inspired genetic algorithm employs a conversion mechanism of quantum representation to code the schedule; and utilizes a rotation angle table to update the population. The effectiveness of the quantum-inspired genetic algorithm is compared to that of a standard genetic algorithm, with experimental results confirming the potential of the proposed approach in tackling complex combinatorial optimization problems.

Keywords: Operations Research, Fuzzy Systems
Abstract: Along with awareness of global warming, there have been many developments in recent years on carbon emissions in transportation. The use of micromobility vehicles has become widespread due to reasons such as energy saving, reducing carbon footprint and reducing traffic density in urban transportation. Although there are useful applications, this new concept has brought with it some problems. It is clear that new regulations are needed on issues such as accessibility, parking areas, security, charging stations, city planning etc. and the big cities need to adapt. Therefore, the carbon neutral districts have been started to be tested in small areas in many parts of the world and their effects are observed. This study performs Selected Element Reduction Approach (SERA), which is a Multi Criteria Decision Making (MCDM) criterion weighting method to determine the weight of the criteria to be used to assess the parking area of e-scooters in carbon neutral areas. SERA is a fuzzy environment MCDM method that aims to give weight to the criterion by the absence effect, which occurs by subtracting the criteria from the general evaluation. In this study, SERA is extended for the first time with single-valued neutrosophic fuzzy sets (SVNFS), which is a three- dimensional fuzzy environment. This study contributes to the literature by analyzing the application of NF-SERA with a numerical example. It also assesses the criteria for the e-scooter parking case for the carbon neutral areas that are currently discussed in the sustainable urban transportation literature.

Keywords: Bio-inspired, Operations Research, Swarm Intelligence
Abstract: This paper presents an innovative approach to addressing the complex challenges posed by Test Task Scheduling Problems (TTSPs) through the utilization of Multimodal Multi-objective Football Game Algorithm (MM-FGA). TTSPs hold significant importance in industries where testing plays a critical role in ensuring product quality. This study outlines the integration of MM-FGA with the Normalized Factor Random Key (NF-RK) encoding scheme, tailored to the discrete nature of TTSPs. Four real-world TTSPs are investigated with the objectives of minimizing makespan and mean workload. Comparative analyses are conducted against other prominent multiobjective algorithms, including NSGA-II, DN-NSGA-II, Tri-MOE-TA&R, MP-MMEA, and MO-Ring-PSO-CD. The results exhibit MM-FGA's competitive performance, particularly in terms of the diversity of solutions obtained in the decision space. This underscores MM-FGA's prowess in addressing the multimodality challenges of optimization problems. The study further suggests the prospect of advancing step-size control strategies through meta-optimization, aiming to refine the algorithm's exploration and exploitation balance for even more potent optimization outcomes. Overall, MM-FGA demonstrates promise in solving multimodal multiobjective discrete problems like TTSPs, while indicating room for future enhancements.

Keywords: Swarm Intelligence, Randomized Algorithms
Abstract: The MKCP (Maximum Set k-Covering Problem) is a widely recognized combinatorial problem that falls under the category of NP-hard problems. It has diverse applications and involves the goal of covering a maximum number of elements using a limited number of candidate sets. In this paper, the novel fixed set search (FSS), a population based metaheuristic, is applied on the problem of interest. The FSS adds a learning mechanism to the greedy randomized adaptive search procedure (GRASP) based on elements frequently occurring in high quality solutions. The main advantage of the proposed approach is the simplicity of implementation compared to the current state-of-the-art methods. The conducted computational experiments show that the FSS even when using a simple local search manages to be highly competitive to state-of-the-art methods. In addition, the FSS manages to find two new lower bounds for the standardly used benchmark test instances. Finally, the performed computational experiments show that the learning mechanism of the FSS significantly improves the performance of the underlying GRASP algorithm.

Keywords: Swarm Intelligence, Particle Swarm Optimization
Abstract: Clustering and classification are the two important tasks involved in pattern recognition. Both tasks are interrelated with each other. The generalization ability of classification learning can be enhanced with clustering results. On the contrary, the class information helps in improving the accuracy of clustering learning. Thus, both learning strategy complements each other. To amalgamate the benefits of both learning strategies, therefore in this paper, we proposed a novel robust kernelized Fuzzy c-Means based multi-objective simultaneous learning framework(RKFCM-MSCC) for both clustering and classification. RKFCM-MSCC employs multiple objective functions to compose the clustering and classification problem, respectively. Both the formulated objective functions are simultaneously optimized using the particle swarm optimization approach. Moreover RKFCM-MSCC uses Bayesian theory that make these multiple objective functions dependent on the single parameter i.e., cluster centers that connect both the clustering and classification learning. The Pareto-optimal solution attained with the RKFCM-MSCC approach complements the clustering and the classification learning process. The effectiveness of the proposed RKFCM-MSCC is empirically investigated on four benchmark datasets and the results are compared with the state-of-the-art approaches.