Search

This article introduces a biologically-inspired model capable of detecting both an object's motion direction and speed, based on retinal neural mechanisms verified through biological experiments. It aims to address the interpretability issues present in current deep learning models. The proposed Motion Detection Neuron (MDN) model, inspired by early research on the retina's direction and speed sensitivity, replicates the motion detection functions of the retina and primary visual cortex. The design of the MDN, inspired by the layered structure of the retina and incorporating various cell types and functions, has been validated through biological experimentation, providing it with robust biological interpretability. Extensive experiments have been conducted to assess the MDN's detection accuracy and robustness against various types of noise. Additionally, to verify that the MDN not only offers enhanced biological interpretability but also maintains detection accuracy comparable to leading deep learning algorithms, we compared its performance with that of LeNet ,EfficientNet and RegNet under identical conditions. The results show that the MDN not only provides better biological interpretability and lower hardware demands but also excels in accuracy under specific conditions, comparable to advanced deep learning algorithms. [Display omitted] • This paper employs a biologically inspired model MDN, which diverges from conventional deep learning approaches by simulating the architecture of the biological retina and the primary visual cortex, thereby providing superior biological interpretability. • The subject of this paper is the detection of object motion, an intrinsic ability of biological organisms, not acquired through learning. Consequently, by employing a direct modeling approach inspired by the biological visual system, this methodology not only circumvents the opaqueness associated with the learning 'black box', but also substantially diminishes both temporal and equipment-related expenditures. • Within the framework of the biologically inspired model, innovatively conducts detection of both an object's motion direction and speed, further advancing the simulation of functionalities inherent to the biological visual system. Furthermore, the experimental scope of this paper extends beyond mere binary images to include grayscale images and color images, and an array of comparative analyses, thereby reinforcing the model's credibility. [ABSTRACT FROM AUTHOR]

In this paper, we explore the possibility of exploiting retrospective clinical data from Electronic Health Records (EHR) for classification tasks in chronic patients. The different intervals, short length and high class imbalance make it unfeasible to use traditional time series techniques. The first contribution of the paper is a preprocessing method to construct a multivariate time series dataset using EHR data, which infers missing data and regularizes the data frequency. The second contribution addresses class imbalance by using domain knowledge and existing short EHR series. We synthetically extrapolate patients' data by using similar long time series and a fuzzy-based approach. The paper addresses the problem of detection of Diabetic Retinopathy (DR). Expert domain knowledge from ophthalmologists has been used in the proposed techniques to guide the processing of time series. The novelty in that case study consists in not using eye-fundus image analysis. Instead, the proposed methods are based solely on EHR data. Several multivariate multiclass time series classifiers are used to detect the four levels of DR severity from the pre-processed data sequences. Experiments prove the quality of the sequence preprocessing techniques proposed for EHR data. Results indicate that the TapNet classifier is the best one for DR grading. Despite being tested for DR detection, the proposed data preparation methods are applicable to other diseases with similar characteristics. [Display omitted] • Historical data from diabetic patients can be used to assess their retinopathy risk. • Long-term diabetic patients are harder to classify than new diabetic patients. • The original sequences have more similarities to the proposed double interpolation. • Data imbalance can be solved by boosting short time series using a fuzzy-based method. • TapNet has the best results among the tested multi-variate time series classifiers. [ABSTRACT FROM AUTHOR]

One of the pivotal challenges in multi-label image classification lies in the fact that each image is often tagged with multiple semantic labels, without the aggregate prediction probabilities being bound to unity. This aspect complicates the straightforward application of conventional single-label image classification algorithms to multi-label contexts. To tackle this challenge, this paper introduces a variational estimation-based multi-label category enhancement fusion distillation technique. The devised loss function focuses on maximizing the biochemical mutual information, thereby enhancing category recognition capabilities. The goal is to adeptly extract and capitalize on the pivotal features of multi-label image scores and structural information, thus elevating the accuracy and efficiency of classification endeavors. This paper not only furnishes a thorough exposition of the issues tackled and the comprehensive architecture of the proposed algorithm but also delineates its operational principles and design rationale via an exhaustive analysis of each critical step within the algorithm. Through an array of experiments across diverse network architectures and datasets, coupled with comparative analyses against extant models and empirical validations, this paper unequivocally validates the efficacy of the suggested algorithm and markedly augments the performance of multi-label classification tasks. [ABSTRACT FROM AUTHOR]

In this paper, bipartite synchronization for memristive neural networks with multi-delay couplings is investigated. The evaluation index of unbounded coupling delays on synchronization could be quantitatively analyzed by considering proportional delay, which will undoubtedly and strongly impede synchronous behavior. By simultaneously taking synchronization and anti-synchronization patterns into account, a novel impulsive controller with a signed form is elaborately designed. For the purpose of selecting suitable impulsive instants, a dynamic self-triggered mechanism is introduced. Additionally, to mitigate the possible risk of the dynamic mechanism transitioning into a static mechanism in exceptional scenarios, a memory-based adaptive updating law is therefore proposed in this paper. It should be noted that the adaptive control related dynamic parameters considered in this paper are in a non-monotonic form. By utilizing Lyapunov stability theorem, parameter variation approach and contradiction analysis method, sufficient conditions for ensuring the synchronization are successfully derived. Finally, two experiments are presented to demonstrate the practicability of the derived results. • Synchronization patterns and anti-synchronization patterns in some previous works could be thought as one of the special cases in our paper. • A non-monotonic dynamic updating law is devised by incorporating memory information, specifically utilizing the sparse historical states of nodes. • Lower bound of dynamic parameter is proved to be greater than a specific positive scalar in this work, that is, the importance of the role played by the dynamic parameter in determining impulsive intervals is enhanced. [ABSTRACT FROM AUTHOR]

Drug recommendation is one of the most crucial research topics in smart healthcare. Its goal is to provide a set of safe drug combination based on the patient's electronic health records (EHRs). Drug recommendation is challenging because it is difficult to obtain an appropriate representation of patient's health state from these personalized historical records. Meanwhile, drug recommendation must prioritize the safety of drug combination because drug–drug interactions (DDIs) could result in side effects. To address these issues, this paper proposes a novel recurrent aggregation neural network for safe drug recommendation, called RASNet. RASNet introduces a straightforward but efficient recurrent aggregation neural network to capture historical records related to the patient's health state of the current visit, which could improve the performance of EHR-based personalized modeling, particularly in cases where the patient's condition changes periodically. Furthermore, this paper presents a novel exponential controller for DDI loss to enhance the safety of drug combination. The proposed controller not only balances the DDI rate between the safety and accuracy of the drug recommendation but also ensures the performance even when the DDI rate is low. Extensive experiments on the MIMIC-III dataset demonstrate that RASNet achieves state-of-the-art performance. Moreover, RASNet exhibits excellent efficiency and safety in drug recommendation. [Display omitted] • Drug recommendation aims to suggest effective and safe drugs based on the patients' medical history records. • RASNet could address the noisy data problem caused by periodic changes due to chronic diseases. • The exponential controller of drug–drug interaction loss could ensure the safety and accuracy of drug recommendation. • RASNet demonstrates outstanding accuracy and efficiency in drug recommendation. [ABSTRACT FROM AUTHOR]

Constrained multi-objective optimization problems (CMOPs) are widespread in practical applications such as engineering design, resource allocation, and scheduling optimization. It is high challenging for CMOPs to balance the convergence and diversity due to conflicting objectives and complex constraints. Researchers have developed a variety of constrained multi-objective optimization algorithms (CMOAs) to find a set of optimal solutions, including evolutionary algorithms and machine learning-based methods. These algorithms exhibit distinct advantages in solving different categories of CMOPs. Recently, constrained multi-objective evolutionary algorithms (CMOEAs) have emerged as a popular approach, with several literature reviews available. However, there is a lack of comprehensive-view survey on the methods of CMOAs, limiting researchers to track the cutting-edge investigations in this research direction. Therefore, this paper reviews the latest algorithms for handling CMOPs. A new classification method is proposed to divide literature, containing classical mathematical methods, evolutionary algorithms and machine learning methods. Subsequently, it reviews the modeling and algorithms of CMOPs in the context of practical applications. Lastly, the paper gives potential research directions with respect to CMOPs. This paper is able to provide guidance and inspiration for scholars studying CMOPs. [ABSTRACT FROM AUTHOR]

Medical image diagnosis has developed rapidly under the impetus of the deep network. Previous works mainly focus on improving the diagnostic accuracy of models, i.e., first use a backbone network to extract image global features and then feed it into the classifier for diagnosis. However, these methods do not fully explore the transparent and reasonable decision-making process of the final classification results, which is crucial for medical diagnosis. In this paper, we propose a framework called Ca usal I ntervention-based M ulti-head A ttention network (CaIMA) to enhance the explainability of medical diagnosis from a causal inference perspective, by exploring the inherent causal relationship between multi-region attention and diagnosis results. Specifically, it consists of three key components: (1) The multi-region attention module enables the network to focus on the distinct discriminative lesion regions that hold causal relationships with the predicted outcome. (2) The attention-driven data augmentation module provides accurate localization of discriminative regions and enhances model explainability. (3) The causal intervention module aims to explore the intrinsic causal relationship between the attention map and the predicted outcome, encouraging the network to learn more useful attention maps for medical image diagnosis. Besides, to address the learning difficulty of this network, we further introduce a non-overlapping multiple attentional guidance loss that encourages the learned multiple attention maps to focus on specific lesion regions without overlapping. We compare the proposed CaIMA with state-of-the-art methods on multimedia medical datasets, including three public medical image datasets (Kvasir, ISIC2018, COVID-19) and one private dataset (CLC), and the experimental results substantiate the effectiveness of CaIMA in terms of diagnosis accuracy and explainability. • This paper provides a fresh perspective on medical image diagnosis from a causal standpoint. • The proposed CaIMA incorporates causal interventions with the multi-region attention framework. • A composite loss is proposed to provide reliable causal visual explanations and enhance model performance. • The effectiveness and superiority of the proposed models are verified. [ABSTRACT FROM AUTHOR]

In the realm of object detection, traditional domain adaptive object detection (DAOD) methods assume that source and target data completely share one identical class space, which is often difficult to satisfy in many real-world applications. To address this limitation, this paper introduces universal domain adaptive object detection (UniDAOD), a learning paradigm that relaxes identical class space assumption to be a different but overlapped class space. Intuitively, the main challenge of UniDAOD is to reduce the negative transfer of private classes (i.e., classes only existed in one domain) and reinforce the positive transfer of the common classes (i.e., classes shared across domains). In this paper, we provide a rigorous theoretical analysis and induce a new generalization bound of the expected target error under the UniDAOD setting. On the basis of theoretical insight, we then propose weighted adaptation (W-adapt) to suppress the interference of private classes and reinforce the positive effects of common classes. In particular, we propose a pseudo category margin (PCM) to quantify class importance based on dynamic pseudotarget label prediction to recognize common classes. Furthermore, to alleviate the impact of inaccurate pseudotarget labels, we propose a temporary memory-based filter (TMF) to dynamically store and update the PCM during progressive training. On the basis of the learned TMF, we design a weighted classwise domain alignment loss to adapt two domains across common classes. Experiments on four universal scenarios (i.e., partial-set, open-partial-set, open-set, and closed-set) show that W-adapt outperforms several domain adaptation methods. [ABSTRACT FROM AUTHOR]

Contrastive learning predicts whether two images belong to the same category by training a model to make their feature representations as close (positive samples) or as far away as possible (negative samples). Selecting appropriate samples is critical to effectively train a model, however, existing methods suffer from false or uninformative sample problems. This paper rethinks how to mine samples in contrastive learning and the proposed method is more comprehensive. It takes into account both positive and negative samples, and mines potential samples from two aspects. First, for positive samples, this paper incorporates both the augmented sample views and the mined sample views. A weighted combination of these positive samples is achieved by using both hard and soft weighting strategies simultaneously. Second, considering the existence of false and uninformative negative samples, this paper analyzes the negative samples from the perspective of gradient and mines negative samples that are neither too difficult nor too easy as potential negative samples, i.e., those negative samples that are close to positive samples. Compared with previous state-of-the-art self-supervised methods, experiments show the obvious advantages of the proposed method, and the corresponding top-1 accuracies of linear classification are improved by 0.77%, 2.39%, and 1.01% on CIFAR10, CIFAR100, and TinyImageNet, respectively. Source code and pretrained models are available at https://github.com/dhkdhk/PSM. [ABSTRACT FROM AUTHOR]

Liver diseases represent a significant healthcare challenge, impacting millions globally and posing complexities in diagnosis. To address this global health concern, this paper introduces a groundbreaking enhancement to the Kepler Optimization Algorithm, termed I-KOA, designed specifically for feature selection in high-dimensional datasets. By harnessing the synergies of Opposition-Based Learning and a Local Escaping Operator grounded in the k-nearest Neighbor (kNN) classifier, I-KOA asserts itself as a potent tool for local exploitation, balanced exploration, and evasion of local optima. To our knowledge, this is the first work to exploit KOA as a feature selection method. Pioneering the utilization of KOA as a feature selection method, the paper rigorously tests I-KOA in two extensive experiments, tackling the complex CEC'22 benchmark suite functions and the intricate landscape of five liver disease datasets. Results underscore I-KOA's unparalleled performance, validated through the Friedman test, where it surpasses seven rival optimization algorithms. Achieving an outstanding overall classification accuracy of 93.46%, Feature selection size of 0.1042, sensitivity of 97.46%, precision of 94.37%, and F1-score of 90.35% across the liver disease datasets, I-KOA's randomized algorithm ensures robust feature selection, striking a compelling balance between subset size and classification efficacy. Acknowledging computational demands and generalization nuances, I-KOA is a formidable tool ready to revolutionize medical diagnosis and decision support systems. The open source codes of the proposed I-KOA are available at https://www.mathworks.com/matlabcentral/fileexchange/161376-improved-kepler-optimization-algorithm. • This paper encompasses a proficient I-KOA algorithm based on OBL and LEO methods. • A new optimized feature selection model for liver disease classification using five datasets. • We employ comprehensive analysis metrics to thoroughly assess the I-KOA algorithm's efficacy. • I-KOA stands out by surpassing its competitors, attesting to its remarkable performance. [ABSTRACT FROM AUTHOR]

In graph few-shot learning, few-shot node classification (FSNC) at the node-level is a popular downstream task. Previous FSNC methods primarily rely on meta-learning or metric learning techniques, aiming to mine prior knowledge from the base classes. However, these methods still have some limitations that need to be addressed, namely: (1) conducting multiple tasks for parameter initialization leads to expensive time costs. (2) ignoring the rich information present in novel classes leads to model over-fitting. To address these issues, this paper proposes a novel graph augmentation method for FSNC on graph data, which includes both parameter initialization and parameter fine-tuning. Specifically, the parameter initialization conducts only one multi-classification task on the base classes, improving generalization ability and reducing time costs. The parameter fine-tuning is designed to include two data augmentation modules (i.e. , support augmentation and shot augmentation) on the novel classes to mine the rich information, thus alleviating model over-fitting. As a result, this paper introduces the first graph augmentation method for FSNC. Experimental results showed that our method achieves supreme performance, compared with state-of-the-art FSNC methods. • We use data augmentation and GCN with optimized initialization for efficiency. • The parameter initialization module aims to improve generalization and efficiency. • The proposed support and shot augmentation modules aim to mine rich information. [ABSTRACT FROM AUTHOR]

This paper focuses on the development of a knowledge-based system for automatically diagnosing issues in Vehicular ad hoc networks (VANETs). VANETs enable communication between vehicles and infrastructure, enhancing road safety and efficiency through timely information exchange. The proposed system aims to efficiently maintain and ensure the continuity of network service by leveraging innovative pattern recognition methods tailored to VANETs. The automatic diagnosis problem in VANETs involves estimating the operating class of network components based on sensor observations. This entails associating sensor measurements with specific operating modes. By implementing condition-based preventive maintenance procedures, potential component failures can be detected early, mitigating network disruptions. Various approaches, such as expert systems, fault trees, network state models, and statistical learning through pattern recognition, can be employed to address this problem. This paper primarily focuses on the statistical learning approach, where a classification or regression function is learned from a set of examples to assign operation modes to new measurements. It discusses relevant metrics and preprocessing techniques to simplify the decision-making process. The diagnostic system's results are determined based on the formulation of the classification or regression problem. The learning base is constructed, and an appropriate classification method is selected to develop and validate the automatic diagnosis system. While non-parametric models like support vector machines are commonly used, this article emphasizes the significance of considering assumptions and leveraging additional information to enhance performance. It proposes a more specific formalization of the problem, integrating the unique characteristics of VANETs. The contributions of this article revolve around the theory of belief functions, a generative approach, and the utilization of parametric models defined using graphical models. Experimental studies conducted on artificial datasets have demonstrated the benefits of the semi-supervised approach within the context of VANET networks. [ABSTRACT FROM AUTHOR]

The railway system is a prime example of a safety-critical system. Predicting the causes of railway accidents holds immense significance in enhancing railway transportation safety. Previous approaches to railway causation analysis have encountered huge challenges regarding data processing and analytical capabilities. To address this concern, this paper proposes an innovative deep model framework based on the Transformer architecture that utilizes historical data on railway equipment accidents to predict the causes behind such incidents. Firstly, this paper proposes the utilization of Convolutional Block Attention in the domain of text processing, serving as a lexical encoder to augment word semantics acquisition in accident texts. Subsequently, in order to address the deficiency of traditional Transformers that lack positional representation information, we propose incorporating a BiGRU (Bidirectional Gated Recurrent Unit) as a contextual positional information encoder to capture contextual positional information in railway accident data effectively. Finally, considering that accident data reports are discrete tabular data, this study suggests employing cue word techniques for preprocessing accident data to alleviate the model's learning burden. We applied the proposed model to the FRA (Federal Railroad Administration) dataset. The results demonstrate that our model surpasses the current state-of-the-art language models, exhibiting superior performance compared to the optimal model with a notable improvement of 3.56%, 0.42%, and 0.76% in Precision, Recall, and F1-score, respectively. Furthermore, our model accurately predicts accident categories prone to misjudgment even when trained on limited data, outperforming existing language models. The study findings will contribute to the prevention and management of railway accidents. [ABSTRACT FROM AUTHOR]

When a multi-fingered dexterous hand interacts with the external environment, it encounters various challenges, including the utilization of complex control techniques and the intricate coordination of finger motion sequences. Previous studies have primarily concentrated on investigating the interaction between multi-fingered dexterous hands and external objects, usually using model-based control or model-free reinforcement learning techniques. However, during practical implementation, various constraining factors are encountered, such as intricate modeling and limited interaction capabilities. In practical scenarios, the utilization of multi-fingered dexterous hands is imperative for the swift and efficient execution of a wide range of interactive tasks, including but not limited to throwing a ball and playing rock-paper-scissors. These tasks require skilled manual dexterity to demonstrate both precise control and quick responsiveness. To tackle this issue, we propose a hierarchical control approach for multi-fingered dexterous hands with interactive functionalities, utilizing model-free deep reinforcement learning. The complex interaction task is decomposed into simple sub-tasks using hierarchical strategy and action primitive decomposition, which effectively reduces the complexity of the action space, and achieves the motion planning and end finger trajectory control of dexterous hand. In a simulated environment, the aforementioned method has successfully executed interactive tasks, including ball throwing and playing rock-paper-scissors. It achieved a maximum normalized reward of 0.83 and an 84 % success rate. These results are noteworthy in terms of both control accuracy and response speed. This study offers novel insights into the effective resolution of the intricate challenges associated with interactions involving multi-fingered dexterous hands and human-computer interaction. [ABSTRACT FROM AUTHOR]

Many real-world datasets encounter the issue of label noise (LN), which significantly degrades the learning performances of classification models. While ensemble learning (EL) has been widely employed to tackle this problem, the Dynamic Selection (DS) of classifiers, as a promising EL branch, is particularly sensitive to LN. To address this issue, a meta-learning-based sample discrimination (MSD) framework is proposed in this paper. Initially, this paper analyzes how LN affects the performance of DS methods through a visual example. Subsequently, under the premise that DS methods are only applicable to samples whose neighborhood is minimally affected or unaffected by LN, a meta-learning dataset is generated in the framework, where the meta-features and meta-labels are derived from the characteristics and the real class distribution of local regions of the samples, respectively. With this dataset, a meta-learner is constructed to determine the feasibility of using DS methods directly to classify a given sample in the presence of LN. For samples that DS methods cannot handle, a novel DS process based on the Genetic Algorithm is designed to mitigate the negative impact of LN. The effectiveness of the MSD framework is validated through extensive experiments conducted on thirty real datasets. These experiments demonstrate the capability of the MSD framework to improve the performances of DS methods across different levels of LN. Furthermore, the efficacy of the proposed MSD framework in handling LN is also highlighted by comparing it with a state-of-the-art method and four mainstream EL methods. [ABSTRACT FROM AUTHOR]

Fuzzy Cognitive Mapping (FCM) and the extensive family of models derived from it have firmly established their strong position in the landscape of machine learning algorithms. Specifically designed for pattern classification and multi-output regression, the recently introduced Recurrence-aware Long-term Cognitive Network (r-LTCN) model is one of these FCM-inspired extensions. On the one hand, this recurrent neural network connects all temporal states generated during the reasoning process with the decision-making layer. On the other hand, it uses a quasi-nonlinear reasoning rule devoted to avoiding convergence issues caused by unique fixed points, which typically emerge in other FCM models. In the original paper, the authors employed a combination of unsupervised and supervised learning to compute the r-LTCNs' learnable parameters. Despite r-LTCNs' astounding performance for a wide variety of pattern classification problems, the literature reports no attempt to train these recurrent neural systems in a fully supervised manner nor provide insights into their performance in other machine learning settings. This paper brings forward a modified Backpropagation Through Time learning (BPTT) algorithm devoted to training r-LTCN models used for multi-output regressions tasks rather than pattern classification. The proposed BPPT includes a simple yet effective mechanism to deal with the vanishing gradient within the recurrent layer that operates as a closed system while being tailored to the quasi-nonlinear reasoning mechanism. Empirical evaluation of the proposed BPTT algorithm using 20 multi-output regression problems reveals that it produces lower prediction errors compared with other state-of-the-art learning approaches. [ABSTRACT FROM AUTHOR]

Zero-shot retrieval aims to transfer knowledge from seen classes to unseen classes by embedding semantic information on class attributes, thus solving the unseen class retrieval problem. However, existing works have focused mainly on unimodal zero-shot retrieval tasks. In this paper, we introduce an efficient method, termed zero-shot discrete hashing with adaptive class correlation (ZSDH-ACC), to speed up cross-modal retrieval. Specifically, this proposed method combines label information with class attribute information to construct a semantic enhancement embedding, in which the problem of class attribute correspondence of multilabel instances can be solved. Furthermore, we learn semantic enhancement embedding to merge more semantic information for feature representation, and its goal is to learn more discriminative hash codes and hash functions. In addition, our proposed method adaptively learns the correlation between class attributes and then embeds more class attribute information into hash codes. Finally, pairwise similarity is used to constrain the learning of hash codes, and thus more discriminative hash codes can be generated. Extensive experimental results on four benchmark multimodal datasets demonstrate that the proposed ZSDH-ACC method can achieve encouraging performance in cross-modal retrieval tasks. The source code of this paper can be obtained from https://github.com/szq0816/ZSDH_ACC. [ABSTRACT FROM AUTHOR]

Evolutionary Algorithms (EAs) are a well-known domain within Artificial Intelligence. EAs have demonstrated their ability to tackle intricate optimization problems using evolutionary theory principles. However, balancing the dual exploration and exploitation processes remains a crucial concern. This paper introduces the Balanced Hill Climbing Weight Algorithm with Diversity (BHWEAD), an innovative approach that combines elements from classic Genetic Algorithm and Differential Evolution. BHWEAD uniquely employs the Hill Climbing local search to guide the influence of its operators, ensuring an optimal interplay between exploration and exploitation. Additionally, it incorporates a diversity control mechanism, resetting specific solutions to prevent premature convergence to suboptimal solutions. The main contribution of the BHWEAD is the mechanism that permits the balance of the exploration and exploitation stages; also, the incorporation of Hill Climbing permits a proper balance of the influence of the operators. Notice that the proposal can escape from suboptimal solutions using a diversity-based strategy. Tested against the CEC2017 benchmark functions in both 50 and 100 dimensions, BHWEAD outperformed 12 notable EAs, underscoring its potential for high-dimensional optimization problems. Besides, the proposed BHWEAD has also been tested over seven engineering problems, and the comparisons include some memetic algorithms., The paper provides additional insights into the algorithm's design, conducts a comparative analysis, and identifies potential areas for improvement. • Introduce a new algorithm that balances exploration and exploitation. • Use easy operators and a GA structure to create a new optimization algorithm. • Use the Hill Climbing to handle the influence of operators in an EA. • Implement a diversity control strategy to avoid falling into suboptimal solutions. [ABSTRACT FROM AUTHOR]

Multi-view clustering has attracted substantial attention thanks to its ability to integrate information from diverse views. However, the existing methods can only generate hard or fuzzy partitions, which cannot effectively represent the uncertainty and imprecision when facing objects in overlapping clusters, thus increasing the risk of error. To solve the above problems, in this paper, we propose an adaptive weighted multi-view evidential clustering (WMVEC) method based on the theory of belief functions to characterize the uncertainty and imprecision in cluster assignment. Technically, we integrate view weight assignments and credal partition between objects and cluster prototypes into a joint learning framework. The credal partition offers a more comprehensive insight into the data by enabling objects to be associated with not only singleton clusters but also subsets of these clusters (termed meta-clusters) and the empty set, which represents a noise cluster. To avoid the interference of irrelevant and redundant features, we further present a weighted multi-view evidential clustering with feature preference (WMVEC-FP) to learn the importance of each feature under different views. We suggest the objective functions of WMVEC and WMVEC-FP and design alternating optimization schemes to obtain the optimal solutions, respectively. Through an extensive array of experiments, it has been demonstrated that our proposed clustering methods outperform other related and state-of-the-art methods in terms of their advantages and overall effectiveness. • The paper presents a multi-view version of evidential clustering with view-weight learning. • The paper further proposes a multi-view evidential clustering with view-weight and feature-weight learning. • The methods create credal partition to represent uncertainty and imprecision in cluster assignment in multi-view data. • Extensive experiments show the better performance of the proposed methods. [ABSTRACT FROM AUTHOR]

Accurate pedestrian trajectory prediction is essential for the advancement of intelligent robot or autonomous vehicle, which is a challenging and interesting task. In this paper, a Scene-Interaction fusion Transformer (SIF-TF) for trajectory prediction is proposed, which takes into account three fundamental factors, i.e. social interaction, past trajectory, and semantic scene. A scene-social modeling method is added to the model to integrate social interaction and semantic scene. The proposed SIF-TF contains two critical components: the scene-social transformer and the temporal transformer. The scene-social transformer is tasked with capturing social interaction and semantic scene information, while the temporal transformer focuses on extracting temporal correlation information. Furthermore, the SIF-TF employs a two-stage trajectory prediction approach to jointly generate future trajectories. To evaluate the effectiveness, the comparative experiments were conducted on five widely-used public datasets. The experiments results, with an average evaluation metric of ADE/FDE of 0.23/0.47, significantly outperforms other state-of-the-art methods. These findings demonstrate that the proposed SIF-TF is capable of delivering more precise pedestrian trajectory predictions across diverse scene backgrounds. [Display omitted] • A Scene-Interaction fusion Transformer (SIF-TF) for trajectory prediction is proposed. • Social interaction information and scene semantic information are effectively fused in the paper. • Temporal correlation, scene semantics and social interaction of a trajectory are taken into account. • As for a specific scenario, a highly robust trajectory prediction model can be obtained via training with only a small amount of data through SIF-TF. • An average performance of 0.23/0.47 on ADE/FDE metrics is achieved by the proposed SIF-TF in the experiments on the five widely-used public datasets. [ABSTRACT FROM AUTHOR]

• This paper extends two statistics which were used in the validation of binary-attribute Q-matrix, for validating the polytomous-attribute Q-matrix. • Based on the two statistics, the paper proposes two algorithms applicable for real-world scenarios with intensive studies to evaluate the performance of the statistics. • Plug in the proposed algorithms, the statistics were compared under various conditions. Guidance on how to validate polytomous-attribute Q-matrix in different scenarios were provided. Compared with typical binary attributes, polytomous attributes can take three or more values (corresponding to different levels of mastery of a respondent or measurement of an item). They can indicate whether a respondent possesses the attributes of interest and mastery levels. Therefore, the test with polytomous-attribute Q -matrix can become more informative and provide respondents with richer diagnostic information than the test based on the dichotomous-attribute Q -matrix. This paper extends the S -statistic and the residual method applicable for the Q -matrix of binary attributes to validate the polytomous-attribute Q -matrix. Under two common scenarios in real-world applications, two associated validation algorithms: the joint validation (JV) algorithm and the online validation (OV) algorithm, are proposed. Both simulation studies and an empirical data example were employed to assess the robustness and usefulness of these two methods under various conditions. Results indicate that the JV algorithm is suitable for validating a Q -matrix predefined by subject matter experts. Especially when the Q -matrix contains fewer misspecifications, while the OV algorithm can be applied to define the attribute vector of "new items". Based on a certain number of "operational items", the OV algorithm can achieve a promising performance for obtaining the specification of the new items. [ABSTRACT FROM AUTHOR]

This paper presents a method for providing explainability in the integration of artificial intelligence (AI) and data mining techniques when dealing with meteorological prediction. Explainable artificial intelligence (XAI) refers to the transparency of AI systems in providing explanations for their predictions and decision-making processes, and contribute to improve prediction accuracy and enhance trust in AI systems. The focus of this paper relies on the interpretability challenges in ordinal classification problems within weather forecasting. Ordinal classification involves predicting weather phenomena with ordered classes, such as temperature ranges, wind speed, precipitation levels, and others. To address this challenge, a novel and general explicable forecasting framework, that combines inductive rules and fuzzy logic, is proposed in this work. Inductive rules, derived from historical weather data, provide a logical and interpretable basis for forecasting; while fuzzy logic handles the uncertainty and imprecision in the weather data. The system predicts a set of probabilities that the incoming sample belongs to each considered class. Moreover, it allows the expert decision-making process to be strengthened by relying on the transparency and physical explainability of the model, and not only on the output of a black-box algorithm. The proposed framework is evaluated using two real-world weather databases related to wind speed and low-visibility events due to fog. The results are compared to both ML classifiers and specific methods for ordinal classification problems, achieving very competitive results in terms of ordinal performance metrics while offering a higher level of explainability and transparency compared to existing approaches. [ABSTRACT FROM AUTHOR]

In recent years, federated learning has made significant progress in preserving data privacy. In this paradigm, clients train local models without sharing their raw data, thereby substantially mitigating the vulnerability to private data exposure. However, it is still possible to infer clients' raw data by leveraging the gradient parameters exchanged between the clients and the server. To address this problem, this paper proposes a novel algorithm that introduces deep compressed sensing into federated learning to support one time encryption, called FL-OTCSEnc, to secure the communication data exchanged between the clients and the server. The process starts by creating a dataset of deep learning model parameters and training a system for both encryption and decryption using deep compressed sensing. This system is then used to secure the communication between clients and the server in federated learning, by encrypting and decrypting the data exchanged. To enhance the security of the proposed algorithm, we introduce an assessment method for evaluating the security level of the clients, facilitating the selection of suitable candidates for deployment within distributed training encryption and decryption models that are updated in real time. To enhance the accuracy of the decrypted deep network model, we introduce a tandem loss function in the training process. Moreover, this paper proves that the proposed end-to-end encryption method satisfies additive homomorphic encryption properties. Extensive experiments demonstrate that the deep compressed sensing encryption in federated learning achieves promising results without increasing the computational complexity. [ABSTRACT FROM AUTHOR]

The Intelligent Transport System (ITS) is an emerging paradigm that offers numerous services at the infrastructure level for vehicle applications. Vehicle-to-infrastructure (V2I) is an advanced form of ITS where diverse vehicle services are deployed on the roadside unit. V2I consists of distributed computing nodes where transport applications are parallel processed. Many research challenges exist in the presented V2I paradigms regarding security, cyber-attacks, and application processing among heterogeneous nodes. These cyber-attacks, Sybil attacks, and their attempts cause a lack of security and degrade the V2I performance in the presented paradigms. This paper presents a new secure blockchain framework that handles cyber-attacks, as mentioned earlier. This paper formulates this complex problem as a combinatorial problem, encompassing concave and convex problems. The convex function minimizes the given constraints, such as time and security risk, and the concave function improves performance and accuracy. Therefore, numerous constraints, such as time, energy, malware detection accuracy, and application deadlines, require optimization for the considered problem. Combining the jointly non-dominated sorting genetic algorithm (NSGA-II) and long short-term memory (LSTM) schemes is the best way to meet the problem's limitations. In this study, the paper designed a malware dataset with known and unknown malware. The different kinds of malware lists (e.g., cyber-attacks) are considered in the form of known and unknown malware lists with the characteristics, size of code, where malware comes from, attack on which data, and current status of the workload after being attacked by the malware. Our main idea is to present blockchain, NSGA-II, and LSTM schemes that handle phishing, routing, Sybil, and 51% of cyber-attacks without compromising application performance. Simulation results show that the study reduces delay and energy, improves accuracy, and minimizes security risks for vehicular applications. [ABSTRACT FROM AUTHOR]

Symbolic data analysis data has provided several advances in regression models concerning the type of symbolic variable. Due to the advantages of using symbolic polygonal data, this paper introduces a linear regression approach for polygonal data based on the generalize linear model theory that provides a unified method to broad range of modeling problems for different types of response as asymmetric continuous and discrete. Ordinary polygonal residuals and a way for finding model inadequacies are presented. Moreover, a quality measure of fit for polygons is also proposed in this paper. Experimental evaluation results illustrate the usefulness of the proposed approach regarding synthetic and real polygonal data. • An approach based on Generalized Linear Models for symbolic polygonal data is proposed. • Polygonal residuals are defined for evaluating the adequacy of the fitted model. • The prediction quality is measured by a metric based on Euclidean distance and polygon vertices. • Synthetic and real polygonal data sets are considered in the experimental evaluation. [ABSTRACT FROM AUTHOR]

An Electrical Submersible Pump (ESP) is an important equipment used in the industry for lifting liquids in various types of wells. An ESP is widely used in the oil industry for offshore exploration. Detecting a faulty ESP before installation is a predictive maintenance measure in order to extend its operational time. Machine learning fault diagnosis is an effective way for performing this task. Machine learning fault diagnosis algorithms are highly dependent of the availability of an appropriate problem dataset. This paper describes in detail the problem of ESP fault diagnosis and the ESPset dataset, a real-world and public dataset for vibration-based fault diagnosis of electrical submersible pumps used on offshore oil exploration. In addition, the paper also proposes an experimental framework for adequately comparing research works based on the ESPset dataset and defines benchmark classifiers and respective results as referential to the fault diagnosis research community. The framework considers the fact that some subset of samples are not drawn independently, and therefore, proposes a cross-validation sampling strategy that mitigates the similarity bias among samples. Indeed, this work shows that a conventional k-fold cross-validation may lead to a clear overestimation of the average performance. This fact is supported by results which show that the best classification model drops from a mean F-measure of 0.887 to 0.733 when removing the similarity bias from the data. • The Electrical Submersible Pumps (ESP) fault diagnosis problem description: a broader and detailed specification. • The ESPset vibration dataset is published: 6032 instances of ESP vibration tests. • An experimental framework to avoid dependence between the training and test phases of the learning task. • The ESPset benchmark classifiers are presented. • The results of the experimental study using the proposed framework and the benchmark classifiers are shown. [ABSTRACT FROM AUTHOR]

Negative sequential patterns (NSP) focus on non-occurring events and play a role that cannot be replaced by positive sequential patterns (PSP). Considering the repetitive occurrence of sequential patterns in a sequence, repetitive NSP (RNSP) mining captures frequent NSP across different sequences from a database. Those patterns benefit many tasks of transaction services, e.g., fraud detection and medical diagnosis. However, limited studies focusing on mining RNSP are proposed, e.g., e-RNSP and ONP-Miner, and they are devised under strict constraints and are inefficient in practice. To address these issues, this paper proposes a Self-adaptive Nonoverlapping RNSP mining method SN-RNSP to mine nonoverlapping RNSP with the self-adaptive gap between successive elements from transaction sequences, which requires that each element cannot be reused at the same position in occurrences, and the gap value does not need to be specified in advance. First, this paper develops a method that maintains occurrences of pattern candidates via the bitmap structure to capture all repetitive PSP (RPSP), which utilizes the bitmap-based operation to calculate support efficiently. Second, SN-RNSP leverages bitmaps to record the locations of RPSP and RNSP in the database and query the repetition times of corresponding RPSP for the support calculation of RNSP. Conducted on real-world and synthetic datasets, extensive experiments demonstrate that SN-RNSP can discover more patterns with better mining performance than the state-of-the-art RNSP mining algorithms in transaction sequence databases. [ABSTRACT FROM AUTHOR]

Automatic text summarization is to obtain a summary by compressing the text while retaining its important information. Then users can obtain the important content of the text by reading the summary. In the research literatures, the extraction summary method is widely used and is also one type of the main research methods of summary methods. However, this extraction summary method still has some problems. The selection of the initial cluster center has not been carefully determined, and the sentence redundancy summarized is high in articles with complex sentences. In order to solve the above problems, this paper proposes an automatic text summarization method based on improved TextRank algorithm and K-Means clustering. This method combines the improved BM25 model and the TextRank algorithm to calculate the BM25 similarity between sentences and obtain the TR scores of sentences. The TR scores are used to select the initial center of clustering based on similarity difference judgment and maximum judgment. The final summary is obtained by combining the cluster scores and sentence scores. The experimental results show that the proposed method in this paper has better evaluation indicators containing ROUGE-1, ROUGE-2 and ROUGE-L than other comparison algorithms including Lead-3, TextRank and MBM25EMB on the DUC2004 dataset. In conclusion, the proposed method in this paper improves the accuracy of automatic text summarization and reduce the redundancy from documents. [ABSTRACT FROM AUTHOR]

Image-Text Retrieval (ITR) is essentially a ranking problem. Given a query caption, the goal is to rank candidate images by relevance, from large to small. The current ITR datasets are constructed in a pairwise manner. Image-text pairs are annotated as positive or negative. Correspondingly, ITR models mainly use pairwise losses, such as triplet loss, to learn to rank. Pairwise-based ITR increases positive pair similarity while decreasing negative pair similarity indiscriminately. However, the relevance between dissimilar negative pairs is different. Pairwise annotations cannot reflect this difference in relevance. In the current datasets, pairwise annotations miss many correlations. There are many potential positive pairs among the pairs labeled as negative. Pairwise-based ITR can only rank positive samples before negative samples, but cannot rank negative samples by relevance. In this paper, we integrate listwise ranking into conventional pairwise-based ITR. Listwise ranking optimizes the entire ranking list based on relevance scores. Specifically, we first propose a Relevance Score Calculation (RSC) module to calculate the relevance score of the entire ranked list. Then we choose the ranking metric, Normalised Discounted Cumulative Gain (NDCG), as the optimization objective. We apply a metric smoothing method named Smooth-NDCG (S-NDCG) to ITR, which transforms the non-differentiable NDCG into a differentiable listwise loss. Our listwise ranking approach can be plug-and-play integrated into current pairwise-based ITR models. Experiments on ITR benchmarks show that integrating listwise ranking can improve the performance of current ITR models and provide more user-friendly retrieval results. The code is available at https://github.com/AAA-Zheng/Listwise_ITR. • The paper addresses the image-text retrieval task. • We integrate listwise ranking into conventional pairwise-based image-text retrieval. • We propose a RSC module to calculate the relevance scores of the entire ranked list. • We apply a metric smoothing method named S-NDCG to image-text retrieval. • Our approach can be plug-and-play integrated into pairwise-based retrieval models. • Experiments show that integrating listwise ranking can improve retrieval performance. [ABSTRACT FROM AUTHOR]

Dyna-style Model-based reinforcement learning (MBRL) methods have demonstrated superior sample efficiency compared to their model-free counterparts, largely attributable to the leverage of learned models. Despite these advancements, the effective application of these learned models remains challenging, largely due to the intricate interdependence between model learning and policy optimization, which presents a significant theoretical gap in this field. This paper bridges this gap by providing a comprehensive theoretical analysis of Dyna-style MBRL for the first time and establishing a return bound in deterministic environments. Building upon this analysis, we propose a novel schema called Model-Based Reinforcement Learning with Model-Free Policy Optimization (MBMFPO). Compared to existing MBRL methods, the proposed schema integrates model-free policy optimization into the MBRL framework, along with some additional techniques. Experimental results on various continuous control tasks demonstrate that MBMFPO can significantly enhance sample efficiency and final performance compared to baseline methods. Furthermore, extensive ablation studies provide robust evidence for the effectiveness of each individual component within the MBMFPO schema. This work advances both the theoretical analysis and practical application of Dyna-style MBRL, paving the way for more efficient reinforcement learning methods. • This paper makes an in-depth analysis of the monotonicity guarantee for the Dyna-style algorithm in the deterministic environment. • A practical schema called MBMFPO is proposed to improve policy performance in real environments. • Experimental results corroborate that the policy trained with MBMFPO schema outperforms the baseline methods in terms of sample efficiency and asymptotic performance. • Further experiments have been conducted to validate the efficacy of the individual components encompassed within the MBMFPO schema. [Display omitted] [ABSTRACT FROM AUTHOR]

• Supervised centers improve the utilization rate of supervision information. • Supervised centers help to separate extremely small clusters. • Feature weights adjust influence of imbalanced features in different clusters. • FW-SSPCM improves clustering results on data with imbalanced sizes and features. • Local information obtained by bilateral filtering improves the anti-noise ability. The possibilistic c-means clustering (PCM) algorithm improves the robustness of fuzzy c-means clustering (FCM) to noise and outliers by releasing the probabilistic constraint of memberships. The semi-supervised possibilistic c-means clustering (SSPCM) algorithm improves the clustering effect on datasets with imbalanced sizes by introducing a small amount of label information. However, the traditional semi-supervised algorithm still faces the problem of low utilization of supervision information for datasets with large differences in sample sizes. Moreover, the Euclidean distance, which treats features equally, cannot handle feature-imbalanced data. Therefore, this paper proposes a semi-supervised possibilistic c-means clustering algorithm based on feature weights (FW-SSPCM) by introducing the ideas of supervised centers. First, the algorithm introduces the supervised center into the objective function of the SSPCM to improve the utilization rate of supervision information and thus guide the center iteration of small clusters. Second, the feature weighting strategy is introduced in the objective function to adaptively assign feature weights according to the importance of different features in different clusters, thus improving the adaptability of the algorithm to feature-imbalanced datasets. In addition, to improve the robustness of the antinoise effect and retain additional image details, a new image segmentation algorithm based on FW-SSPCM and local information (LFW-SSPCM) is proposed by introducing local spatial information obtained by bilateral filtering. Finally, through clustering experiments on synthetic data, UCI datasets and on color images characteristic of multiple features, including imbalanced sizes, imbalanced features and strong noise injection, the clustering performances of the proposed FW-SSPCM and LFW-SSPCM proposed in this paper are significantly better than those of several related clustering algorithms. [Display omitted] [ABSTRACT FROM AUTHOR]

With the development of image generation and processing techniques, image inpainting techniques based on deep learning have achieved impressive results. Especially the emphasis on global context during inpainting enables the network to generate reasonably coarse inpainting results at low resolutions. However, how to better achieve high-quality texture filling at high resolutions is still a challenging problem. To address this problem, most methods design two-stage networks to achieve structure and texture restoration separately. But in the face of large-scale masks, the generated textures still suffer from blurring and artifacts. Therefore, in order to achieve inpainting of images with large-scale masks and generate fine textures, this paper proposes an end-to-end generative adversarial model for large mask inpainting, called Panoramic Feature Aggregation Network (PFAN). First, this paper designs a Euclidean Attention Mechanism (EAM) which exploits encoder features to generate low-resolution structure restoration. Then a Feature Aggregation Synthesis Block (FASB) is proposed in the decoder to achieve high-resolution complex texture filling. With the global receptive fields of these two modules, texture filling results with satisfactory performance even under large-scale masks. Experiments on CelebA-HQ, Paris Street View and FFHQ datasets show that the proposed method has superior performance. [ABSTRACT FROM AUTHOR]

Reconfigurable intelligent surface (RIS) is expected to enhance task offloading performance in non-line-of-sight mobile edge computing (MEC) scenarios. This paper aims at reducing the long-term energy consumption for task offloading in a MEC system assisted by double RISs while considering non-orthogonal multiple access (NOMA). Due to the non-convexity of the formulated problem, we propose an action-constrained deep reinforcement learning (DRL) framework based on twin delayed deep deterministic policy gradient (TD3) algorithm that includes inner and outer optimization processes, which greatly reduces the action space size of the DRL agent, making it easy to implement and achieve fast convergence. During the inner optimization phase, by utilizing theoretical derivations, we propose a low-complexity method to optimally derive the transmit power, the local computing frequency, and the computation resource allocation in the base station. In the outer optimization phase, building on the solution derived from the inner optimization, we further use the TD3 algorithm to jointly determine the phase shifting of RIS elements, the offloading ratio, and the transmission time for each time slot. Experiment results demonstrate that the proposed algorithm achieves rapid convergence performance. In comparison to single RIS assisted offloading, the double RISs assisted offloading scheme proposed in this paper can reduce energy consumption by 42.8% on average. [ABSTRACT FROM AUTHOR]

Standard k -means clustering necessitates computing pairwise Euclidean distances between each instance x in a data set D and all cluster centers, resulting in inadequate efficiency when dealing with high-dimensional data sets. Given its widespread usage, it is imperative that k -means clustering should be performed quickly to ensure efficient solutions. This paper is dedicated to exploring ways to improve the efficiency of the k -means algorithm in high-dimensional space. Unlike approximated approaches, our proposed method LBKC can achieve acceleration while yielding clustering results that are the same as what standard k -means clustering generates. LBKC utilizes the lower bound of Euclidean distance to safely avoid a large number of unnecessary distance calculations, thus achieving the goal of accelerating k -means process. Three carefully designed lower bounds based on the block vector, segment mean, and nonlinear embedding are presented in this paper, and they are employed in the proposed method. Furthermore, our approach LBKC is orthogonal to state-of-the-art methods, and we show how LBKC can be naturally combined with them to further improve their performance. Comprehensive experiments are conducted on a variety of data sets to evaluate the performance of the proposed approaches and related competitors, and the experimental results verify the effectiveness of our proposals. [Display omitted] • A new framework named LBKC is proposed for accelerating k-means clustering. • Three carefully designed lower bounds are presented and embedded into LBKC. • LBKC can be combined with the state-of-the-art to improve their performance. • We conduct extensive experiments to validate the effectiveness of the proposals. [ABSTRACT FROM AUTHOR]

The unsupervised domain adaptation semantic segmentation task is challenging due to the distribution shift problem between the source and the target domains. In this paper, we provide a novel perspective to address this problem. Currently, in the literature, it is shown that output-level domain adaptation networks (OL-DAN) can generate outputs with smaller distribution shift. Motivated by this phenomenon, a Progressive Domain Adaptation (PDA) framework is proposed, which uses the outputs generated by OL-DAN as the auxiliary domain images to progressively align the distribution between the source and target domains. Unlike existing two-stage input-level domain adaptation methods which use an image translation network as a standalone model to generate the auxiliary domain images, the PDA is an end-to-end framework that contains an OL-DAN and a domain fusion domain adaptation network (DF-DAN). The OL-DAN aims to gradually generate the outputs with smaller distribution shift and more accurate semantic structures as the auxiliary domain images in every iteration optimization. The DF-DAN is proposed to further mine the domain-invariant information from the auxiliary images and then fuse the features learned from the original images and the auxiliary images to obtain a richer representation. Finally, the distribution between the source and target domains is aligned to optimize the OL-DAN and DF-DAN. Experiments demonstrate that the proposed PDA achieves superior performance on three cross-domain semantic segmentation benchmarks. • A progressive domain adaptation framework is proposed for semantic segmentation. • This paper provides a novel perspective to generate auxiliary domain images. • The proposed approach gains competitive performance on three challenging benchmarks. [ABSTRACT FROM AUTHOR]

The shipping industry has experienced rapid growth in recent years, prompting a need for advanced target recognition technology based on marine radar. This paper introduces the Track Classification Model (TCM), a novel approach for classifying track sequences in real scenarios. The TCM utilizes a feature extraction network based on multi-feature fusion, taking radar echo images and motion information of the target as input, to improve classification accuracy. Additionally, the paper also presents a dataset production method that addresses the issue of missing labels, a critical problem in track sequence classification. Through ablation experiments, the paper demonstrates the effectiveness of the design strategy, with the multi-feature fusion network successfully extracting features and achieving superior performance over single feature extraction networks. The results show that increasing the number of input track points and raising the upper limit of the input sequence leads to improved classification accuracy. Finally, in real scenarios, the proposed model outperforms other algorithms, showcasing its high engineering application value. • Using radar echo image information and target motion information as the input of the network, multi-feature fusion extracts features of track characteristics. • A track classification model based on Transformer is designed for track sequences. • The relationship between track classification and input track point length is obtained, resulting in the optimal input sequence length. [ABSTRACT FROM AUTHOR]

Genetic programming (GP) has been considerably used for image classification because of its ability to learn simple and effective models. However, most GP methods require a large amount of training data to learn informative features for classification, where the generalization performance might be poor when only a few training instances are available. In addition to using classification accuracy to assess the goodness of GP individuals/solutions like in most GP methods, this paper proposes a new fitness function containing distance measures. The proposed method uses different distance measures to deal with binary and multi-class classification automatically. By simultaneously minimizing the within-class distance and maximizing the between-class distance, the generalization performance can be improved. Furthermore, existing GP methods typically employ standard crossover to search for the best individuals from the whole search space. However, these methods might not completely exploit the potential local search space. Based on the niching technique, this paper develops a new crossover operator, which enables better exploitation of the global and local search space, improving learning effectiveness and classification accuracy. The new approach achieves significantly better generalization performance than almost all benchmark methods on eight datasets and is also computationally efficient. Further analysis demonstrates the significance of the new fitness function and crossover operator and shows the potentially good interpretability of the learned models. • A GP-based method is proposed for image classification with small training data. • A fitness function is created to evolve well-generalized solutions in the new method. • A new crossover is developed to better explore and exploit the search space of GP. [ABSTRACT FROM AUTHOR]

The Industrial Internet of Things (IIOT) contains many devices from different autonomous domains (e.g., factories), which need to cooperate to complete complex manufacturing process. Devices from different domains collaborate with each other, which greatly raises trust concerns about device-to-device interactions. Existing trust management approaches may result in the risk of privacy leakage and low accuracy of trust evaluation. Thus, trust issues during interaction remain unsolved but imperative. In this paper, we propose a blockchain-based privacy-preserving trust management architecture PPTMA. Specifically, PPTMA adopts federated learning to train a task-specific trust model for different collaborative task. Our work is the first attempt to research the relationship between the weight calculation of trust metric and the change of context in the field of trust management. To preserve the privacy of devices, differential privacy (DP) is exploited during the trust evaluation process. In addition, a game theory-based incentive mechanism is proposed to encourage the IIOT device for actively and honestly submitting the trust data into the blockchain as so to promote the accuracy of trust computing. Finally, we also design a parallel consensus protocol (OPBFT) which realizes an assembly line to speed up the efficiency of the consensus process. The idea of consensus assembly line firstly proposed by us brings new opportunities for improving the consensus efficiency. Extensive experiments have been conducted to show the effectiveness and efficiency of the proposed method. • This paper proposes a privacy-preserving trust management architecture (PPTMA) that integrates blockchain technology. With the block ledger, the devices from different IoT systems can share the trust data in a distributed way, and blockchain also provide a traceable approach where the process of trust evaluating is recorded and verified as the transactions due to blockchain's features. However, different from existing methods directly using blockchain that often ignore to protect the user's privacy information included in the trust data, PPTMA uses differential privacy (DP) to add noise to the data before uploading trust data. • A common problem is that most of the mainstream consensus protocols (PoW, PoS, DPoS, PBFT, et, al.) incur the high time cost with many transactions. This paper also focusses on the hot issue. We utilize the idea of parallel computing to improve the traditional PBFT (Practical Byzantine fault Tolerance) consensus protocol. The proposed new consensus protocol is presented as OPBFT, where the pack and consensus stages are designed to work in parallel. Compared with the traditional PBFT protocol where every round of consensus is completed serially, OPBFT realizes an assembly line to speed up the efficiency of the consensus process. The blocks are formed and confirmed in parallel with OPBFT protocol. The idea of consensus assembly line firstly proposed by us brings new opportunities for improving the consensus efficiency. • Trust is related with context. In the industrial IoT, trustor X may trust trustee Y for one task but not for another one. That is, the trustworthiness of an IIOT device should change as the context changes. different from existing methods rarely considering the impact of context when evaluating the trustworthiness of an IIOT device, our PPTMA leverages federated learning to train a specific trust model for different context. This kind of way not only can be adapted to the change of context, but also it effectively improves the accuracy of trust calculation. Our work is the first attempt to research the relationship between the weight calculation of trust metric and the change of context in the field of trust management. • In industrial internet of Things, the devices are reluctant to actively share their trust data due to the cost issue and privacy issue. However, most current trust management frameworks rarely care about the quality and quantity of trust data, and mainly focus on how to implement an effective trust evaluation algorithm. In fact, the power of the trust management systems relies heavily on the quantity and quality of trust data submitted by IIOT devices. Therefore, it is needed to design an effective incentive mechanism to encourage devices to participate the dissemination of trust data. With the game theory, we innovatively model the process of trust data submitting as a game and analyze the game process between the payment and punishment. In the end, nash equilibrium point is obtained. The designed incentive mechanism by us greatly reduces the possibility of devices uploading false data. [ABSTRACT FROM AUTHOR]

Due to inter-individual variances, cross-subject electroencephalogram (EEG)-based emotion recognition is a challenging task. In this paper, we construct a multi-branch Capsule network (named DA-CapsNet) based on domain adaptation to improve the performance of cross-subject EEG emotion recognition. To fully capture the various intensity characteristics of a single emotion, firstly, DA-CapsNet decomposes the source and the target domain EEG signals into four frequency bands and homomorphically groups the data in each band, and then extracts the differential entropy (DE) features for each group separately. Taking into account the spatial arrangement of the electrodes, the DE features are mapped into a two-dimensional matrix to form a homomorphic difference cube sequence (HDCS). Second, to enhance the feature information of the same emotion and accelerate the run efficiency of the network, a parallel structured multi-branch primary Capsual network (CapsNet) is constructed in this paper. The multi-branch primary CapsNet can effectively extract the aforementioned sequence discriminative features and fuse them as the input features of the capsule emotion classifier. Finally, to lessen inter-domain distribution discrepancies, we brought adversarial domain adaptation to improve the performance of cross-subject emotion recognition. Numerous tests are run on the three public datasets of EEG, and the results show that the proposed algorithm in this paper works well. [ABSTRACT FROM AUTHOR]

Multi-person pose estimation (MPPE) presents a challenging yet crucial task in computer vision. Most existing methods predominantly concentrate on isolated interaction either between instances or joints, which is inadequate for scenarios demanding concurrent localization of both instances and joints. This paper introduces a novel CNN-based single-stage method, named Dual-path Hierarchical Relation Network (DHRNet), to extract instance-to-joint and joint-to-instance interactions concurrently. Specifically, we design a dual-path interaction modeling module (DIM) that strategically organizes cross-instance and cross-joint interaction modeling modules in two complementary orders, enriching interaction information by integrating merits from different correlation modeling branches. Notably, DHRNet excels in joint localization by leveraging information from other instances and joints. We evaluate the proposed DHRNet on the COCO, CrwodPose, and OCHuman datasets. Our DHRNet achieves competitive results on the COCO (69.9% AP) and CrowdPose (71.5% AP) datasets. And DHRNet achieves the current state-of-the-art results on the OCHuman dataset (58.5% AP). Experimental results showcase the effectiveness of DHRNet. The code will be released at https://github.com/YHDang/dhrnet-multi-pose-estimation. • We propose a CNN-based Dual-path Hierarchical Relation Network for multi-person human pose estimation. • We design a dual-path interaction module (DIM) to simutaneously capture joint and instance interactions. • Guided by DIM, the model locates joints by integrating information from other instances and joints. • The method excels on COCO, CrowdPose, and OCHuman benchmarks, proving its effectiveness. [ABSTRACT FROM AUTHOR]

Remote Photoplethysmography (rPPG) technique plays a pivotal role in non-contact heart rate (HR) estimation. The existing deep learning-based technology has obtained better performance than the traditional method. However, there is a large amount of redundant information in the pulse signal due to noise from light, motion, etc. Meanwhile, single region of interest (ROI)-based methods cannot get accurate prediction results when the region is occluded. For this reason, this paper proposes a Bi-level weighted mixed-domain self-attention network (Bi-WMSN) for non-contact HR estimation. In the first-level weighting, ROI adaptive weighted blocks (RAWB) are developed to adaptively fit the weights of signals from different ROI regions. In the second-level weighting, a dual-channel mixed-domain adaptive weighted network (DMAWN) is designed to adaptively suppress redundant information in the input signal and highlight features related to HR estimation. The weights of different feature information are calibrated to improve the performance of HR estimation. Finally, the results of the Bi-level weighting are input into the HR estimation framework to obtain the estimated HR. The proposed method outperforms those state-of-the-art in terms of many evaluation metrics on the two datasets: MAHNOB-HCI (MAE = 2.18, RMSE = 3.17), PURE (MAE = 2.08, RMSE = 3.18). [ABSTRACT FROM AUTHOR]

In recent years, two competitive time series classification models, namely, ROCKET and MINIROCKET, have garnered considerable attention due to their low training cost and high accuracy. However, they rely on a large number of random 1-D convolutional kernels to comprehensively capture features, which is incompatible with resource-constrained devices. Despite the development of heuristic algorithms designed to recognize and prune redundant kernels, the inherent time-consuming nature of evolutionary algorithms hinders efficient evaluation. To efficiently prune models, this paper eliminates feature groups contributing minimally to the classifier, thereby discarding the associated random kernels without direct evaluation. To this end, we incorporate both group-level (l 2 , 1 -norm) and element-level (l 2 -norm) regularizations to the classifier, formulating the pruning challenge as a group elastic net classification problem. An ADMM-based algorithm is initially introduced to solve the problem, but it is computationally intensive. Building on the ADMM-based algorithm, we then propose our core algorithm, POCKET, which significantly speeds up the process by dividing the task into two sequential stages. In Stage 1, POCKET utilizes dynamically varying penalties to efficiently achieve group sparsity within the classifier, removing features associated with zero weights and their corresponding kernels. In Stage 2, the remaining kernels and features are used to refit a l 2 -regularized classifier for enhanced performance. Experimental results on diverse time series datasets show that POCKET prunes up to 60% of kernels without a significant reduction in accuracy and performs 11 × faster than its counterparts. Our code is publicly available at https://github.com/ShaowuChen/POCKET. [ABSTRACT FROM AUTHOR]

The literature on fixed effects panel interval-valued data models has been established. However, less attention has been given to models that simultaneously, consider the interval information of panel data and explore the nonlinear relationship between interval variables. To deal with this issue, this paper formulates a nonparametric fixed effects panel interval-valued data model. To estimate the fixed effects and nonparametric component, we propose a locally linear method based on the profile least squares framework. Later, experiment results on synthetic and real data sets illustrate the advantages of our proposed model. [ABSTRACT FROM AUTHOR]

Deep stochastic configuration networks (DeepSCNs), as a kind of randomized learner model, have the capability to generate a learning representation quickly and efficiently. Based on DeepSCNs, convolutional stochastic configuration networks (ConSCNs) demonstrated superiority in spectroscopic data analytics by utilizing 1D convolutional and pooling operations. However, directly employing 1D ConSCNs in image data analytics could potentially lead to the loss of spatial information in images and poor generalization performance. This paper extends the original ConSCNs to a 2D version, termed 2DConSCNs, aiming at rapidly constructing randomized learners with 2D input shapes. Empirical results on eight benchmark datasets demonstrate the proposed 2DConSCNs outperform several existing randomized learner models and show good potential for image data analytics. [ABSTRACT FROM AUTHOR]

The neighborhood information theory have been used to evaluate the relevance and redundancy in feature selection for mixed data containing discrete and continuous features. However, existing neighborhood information theory-based feature selection methods commonly select the same subset of features for all classes, which may not necessarily be optimal or suitable for every class. This paper proposes a class-specific feature selection method based on neighborhood information theory, which allows the selection of a unique subset of features for each class in the context of mixed data. Firstly, several class-specific neighborhood information measures are introduced, including class-specific neighborhood mutual information (CSNMI), class-specific neighborhood conditional mutual information (CSNCMI), and class-specific neighborhood joint mutual information (CSNJMI). Then, a relevance-redundancy weight is defined based on CSNCMI and CSNJMI to extract relevant and redundant information about a specific class. Furthermore, a new feature selection algorithm is designed based on CSNMI and this relevance-redundancy weight that can select an appropriate subset of features for each class. Lastly, an ensemble framework is presented for integrating the feature subsets selected for each class. The experimental results demonstrate the effectiveness of the proposed class-specific feature selection method. [ABSTRACT FROM AUTHOR]

Graph Neural Networks (GNNs) have recently achieved remarkable success in learning graph structures and have been applied in a variety of practical applications such as medical diagnosis, drug discovery, chemical compound synthesis, and traffic forecasting. However, like other deep learning methods, GNNs are black-box networks with inexplicable results, restricting the applicability of GNNs in many important applications such as biochemistry and medicine. This paper focuses on the field of explaining graph learning methods, which is still less explored and lacks strong theoretical grounds. Based on Shapley value from game theory, we propose a novel graph explainer, named Shapley and Embedding-based Graph Explainer (SEGE), which builds a linear approximator in node embedding space to interpret the GNNs' prediction of an input graph. Experiments on synthetic and real-world datasets demonstrate that SEGE achieves state-of-the-art performance and is significantly better than baseline models both qualitatively and quantitatively. Moreover, SEGE owns the capacity of automorphic equivalence preservation, which is not observed in existing explainers. [ABSTRACT FROM AUTHOR]

Nonlinear equation systems (NESs) are common in practical applications, and solving them is an important task in numerical computation. Evolutionary algorithms (EAs) for handling NESs have received considerable attention. EAs generate a large amount of data, which reflect the evolutionary behavior. However, there are still deficiencies in the mining and use of these data. Inspired by the ability of humans to acquire knowledge from past historical experience, this paper proposes a knowledge-learning-and-transfer-aided hybrid niching-based differential evolution (HNDE/KLT). HNDE/KLT aims to acquire knowledge from historical experiences and use them to guide the evolution. The HNDE/KLT include two main features: i) artificial neural networks are embedded in EAs, so the algorithm can learn from successful historical evolutionary information and obtain the relationship between the current individual's position and the optimal evolutionary direction; ii) a knowledge-transfer-based technique is designed to perform information exchange between different subpopulations, enhancing exploitation efficiency. Experimental results show that the HNDE/KLT can better solve NESs, indicating that learning plus transfer can make problem solving more intelligent. Additionally, we employed our algorithm to solve two real-world problems, and obtained good results. [ABSTRACT FROM AUTHOR]