Multi-agent systems have gained prominence in various applications for their scalability, flexibility, and efficiency. However, generating coordinated formation trajectories in a distributed manner remains challenging due to the lack of global information or designated leaders. To address this issue, we propose a fully distributed framework based on dynamic game theory. The problem is formulated as a dynamic trajectory game with a potential structure, jointly optimizing collision avoidance, control effort, formation coordination, and goal-reaching objectives. To compute the corresponding Nash equilibrium (NE), we introduce a distributed algorithm termed ADRD. By leveraging the consensus Alternating Direction Method of Multipliers (ADMM), our approach decouples inter-agent interactions. This enables each agent to independently compute its trajectory with theoretical convergence guarantees. Extensive simulations in obstacle-rich environments demonstrate that ADRD produces safe and coordinated trajectories with superior robustness compared to centralized and distributed baselines. Quantitative comparisons show consistent improvements over the proximal iterative best response scheme in formation accuracy, control cost, and overall cost metrics. Scalability experiments with up to 12 agents further validate the effectiveness of the method in larger teams.

Deep learning-based incomplete multi-view clustering has gained prominence for clustering tasks due to its superior feature learning capabilities across multiple views. Nevertheless, considering that the data incompleteness significantly weakens the adequate information of multi-view data, existing methodologies often prioritize filling missing data with the average values of observed cross-view features. Yet, they do not consider using the intrinsic structural relationship among diverse views within the same cluster. These approaches may inadvertently introduce noise or irrelevant information, degrading clustering performance when using recovered data. To address these limitations, we propose the Attention-based Deep Incomplete Multi-view Clustering via Bi-Alignment Guidance (ADIMC-BAG). Specifically, we design an attention layer module to enhance the compactness of view-specific sample-prototype relationships and recover the missing data. Besides, we develop a bi-alignment guidance strategy that ensures the learning of prototypes consistency across views and further enables securing more discriminative features and precise cluster assignments. By capturing the compactness of the view-specific sample-prototype relationship and the consistency of cross-view prototypes, ADIMC-BAG can acquire the commonality of within-cluster samples across views, which is conducive to restoring missing data. Experiments on seven multi-view benchmarks demonstrate the method’s effectiveness, advancing incomplete multi-view clustering research and providing a robust solution for real-world scenarios.

Emotion-Cause Pair Extraction (ECPE) aims to identify and extract emotion clauses and their corresponding cause clauses from unannotated emotional texts. Previous approaches primarily leveraged contextual features or attention weights as semantic information to enhance feature representations. However, these methods often overlook the additional causal information that external knowledge can bring to ECPE. Given the evident inferential relationship between emotions and their causes, we propose a novel model called Rhetorical Structure Theory and Inference-Aware Graph networks (RSTIAG) to address the challenges posed by ECPE. This model constructs a semantic dependency space grounded in discourse-level rhetorical structure theory and a symbolic reasoning space based on an inference-aware graph. By integrating the introduced knowledge through a space fusion co-attention mechanism and modeling the dependencies between these two spaces, the model enables clauses to simultaneously capture diverse knowledge from both spaces. Additionally, we enrich the information dimensions of emotion-cause pairs by instantiating the reasoning process from the perspectives of causality and implicit inference. To substantiate the effectiveness of our proposed RSTIAG model, we undertake comprehensive experiments. The results demonstrate that it significantly outperforms 11 baseline models on an English benchmark corpus, confirming its effectiveness.

The detection of Multiple Sclerosis (MS) and Myelitis from MRI brain modalities is a challenging task that requires tailored transformer-based attention mechanisms. This paper presents MS-Trust that features a causality attention block to maintain the sequential integrity of MRI data and a global attention block to capture long-range dependencies and global contextual information, with squeeze-and-excitation block to recalibrate channel-wise feature responses. MS-Trust's iterative architecture emphasizes more informative features while suppressing less useful ones, to focus on key indicators of MS and Myelitis such as lesions or areas of inflammation. MS-Trust, through its unique architecture, improves generalization through CutMix and MixUp regularizations, and a compact convolutional tokenizer to enhance the classification of MS and Myelitis. We conducted ablation studies of the main components that lead to elevated performance. Two datasets with a total of 6173 images comprised axial and sagittal MRI brain modalities that were prospectively used for Myelitis detection in the context of MS. Across all metrics, the proposed MS-Trust tend to outperform the CCT, BiT, BiT-CutMix, and BiT-MixUp benchmarking models, especially in the combined modality, indicating that MS-Trust might be better suited for handling complex data from multiple sources. Our best MS detection can reach accuracies of 90–92% for the axial, sagittal, and both modalities while the benchmarking models do not generally exceed 86%. Comparative analysis for Myelitis detection shows that MS-Trust achieves an approximated accuracy of 88–90% using both modalities while the benchmarking models do not exceed 79%. Furthermore, MS-Trust outperformed no-attention MLP architectures, plus more than 15 pretrained models and vision transformers using transfer learning. Despite many significant efforts and promising outcomes in this domain, accurate classification of Myelitis and MS subtypes remains a challenging task. Further evaluation and validation are needed to assess its generalizability to different datasets and settings.

Deep reinforcement learning (DRL) has been extensively used to address portfolio optimization problems. DRL agents acquire knowledge and make decisions through unsupervised interactions with their environment without requiring explicit knowledge of the joint dynamics of portfolio assets. Among these DRL algorithms, the combination of actor-critic algorithms and deep function approximators is the most widely used DRL algorithm. Here, we find that training the DRL agent using the actor-critic algorithm and deep function approximators may lead to scenarios where the improvement in the DRL agent's risk-adjusted profitability is insignificant. We argue that such situations primarily arise from the following two problems: sparsity in positive reward and the curse of dimensionality. These limitations prevent DRL agents from comprehensively learning asset price change patterns in the training environment. As a result, the DRL agents cannot effectively explore the dynamic portfolio optimization policy to improve the risk-adjusted profitability in the training process. To address these problems, we propose a novel multi-agent learning system based on the hierarchical deep reinforcement learning (HDRL) algorithmic framework in this research. Under this framework, the agents work together as a learning system for portfolio optimization. Specifically, by designing an auxiliary agent that works together with the executive agent for optimal policy exploration, the learning system can focus on exploring the policy with higher risk-adjusted return in the action space with positive return and low variance. The performance of the proposed learning system is evaluated using a portfolio of 29 stocks from the Dow Jones index in four different experiments. In the training process, the objective functions of the actor and critic both ultimately achieve stable convergence in the training process. The risk-adjusted profitability of our learning system in the training environment is significantly improved. Hence, we prove that the policies executed by our learning system in out-sample experiments originate from the DRL agents' comprehensive learning of asset price change patterns in the training environment. Furthermore, we find that adopting the auxiliary agent and HDRL training algorithm can efficiently overcome the issue of the curse of dimensionality and improve the training efficiency in the positive reward sparse environment. In each back-test experiment, the proposed learning system is compared to sixteen traditional strategies and ten strategies based on machine learning algorithms in the performance of profitability and risk control ability. The empirical results in the four evaluation experiments demonstrate the efficacy of our learning system, which outperforms all other strategies by at least 8.2% in terms of Sharpe ratio, Sorino ratio, and Calmar ratio. This indicates that the policies learned in the training environment can exhibit excellent generalization ability in the back-testing experiments.

The purpose of this paper is to propose a new two-stage one-to-two matching decision-making method in a Fermatean fuzzy environment to make up for limitations of reviewer-paper matching research. The research framework is developed in two phases: in the first stage, according to Fermatean fuzzy evaluation matrices of reviewers with respect to the attribute set, the editor of journals uses the grey correlation method to select reviewers. In the second stage, positive and negative ideal points of two-sided agents are calculated to construct positive and negative prospect matrices and income and loss matrices; on this basis, comprehensive positive and negative prospect matrices are obtained. Subsequently, a single-objective programming model based on the idea of maximizing deviation and linear inequality constraints is established to determine attribute weights; then, a positive similarity matrix, a negative similarity matrix, and a fairness matrix are calculated. Furthermore, a one-to-two matching model between papers and reviewers is established, and the optimal matching scheme is acquired by solving the model. Finally, the practicability of the proposed method is verified by an example analysis of reviewer-paper matching, effectively addressing the challenges of information uncertainty and bilateral satisfaction optimization in the review process. Sensitivity testing and comparative analysis further confirm the superiority of the proposed method over conventional ones.

Surrogate-assisted evolutionary algorithms provide an effective means for complex and computationally expensive optimization problems. However, due to the scarcity of training samples, the prediction accuracy of frequently-used regression surrogate models can hardly be guaranteed as the difficulty of the problem increases, resulting in performance degradation of the whole algorithm. Since real-world problems rarely exist in isolation, it is expected to alleviate the above issue by properly exploiting the knowledge shared across different problems. In this context, this study proposes a novel evolutionary multitasking optimization algorithm assisted by a classifier rather than a regression model for expensive multitasking problems, where the accuracy of the classifier is boosted by knowledge transfer. To be specific, a support vector classifier (SVC) is first developed and integrated into a classic evolutionary algorithm, i.e., covariance matrix adaptation evolution strategy (CMA-ES). With a low computational cost, it helps CMA-ES to prescreen parent solutions from the current population. Following that, a knowledge transfer strategy is designed to enrich the training samples for each task-oriented classifier by sharing high-quality solutions among different tasks, where a PCA-based subspace alignment technique is employed. Extensive experiments indicate that the SVC-assisted CMA-ES gains significant superiority over general CMA-ES in terms of both robustness and scalability, and the knowledge transfer strategy further helps it earn a competitive edge over some state-of-the-art algorithms on expensive multitasking optimization problems.

In the Page 11, Line 3 published as "For the HalfCheetah Forward Backward task, we use the optimal data from [30] and an inexpert, partially trained SAC [61] policy to generate suboptimal demonstration data, aiming to evaluate the robustness of our algorithm."It should have read as "For the HalfCheetah Forward Backward task, we use the optimal data from [30] and an inexpert, partially trained SAC [61] policy to generate suboptimal demonstration data."

Emotion recognition in conversations has recently emerged as a hot research topic owing to its increasingly important role in developing intelligent empathy services. Thoroughly exploring the conversational context and accurately capturing emotion-shift information are highly crucial for accurate emotion recognition in conversations. However, existing studies generally failed to fully understand the complex conversational context due to their insufficient capabilities in extracting and integrating emotional cues. Moreover, they mainly focused on the speaker’s emotion inertia while paying less attention to explore multi-perspective emotion-shift patterns. To address these limitations, this study proposes a novel multimodal approach, namely, GAT-CRESA (Graph ATtention based on Contextual Reasoning and Emotion-Shift Awareness). Specifically, the multi-turn global contextual reasoning module iteratively performs contextual perception and cognitive reasoning for efficiently understanding the global conversational context. Then, GAT-CRESA explores emotion-shift information among utterances from both the speaker-dependent and the global context-based perspectives. Next, the emotion-shift awareness graphs are constructed for extracting significant local-level conversational context, where edge relations are determined by the learnt emotion-shift labels. Finally, the outputs of graphs are concatenated for final emotion recognition. The loss of emotion prediction task is combined together with those of two perspective’s emotion-shift learning for guiding the training process. Experimental results show that our GAT-CRESA achieves new state-of-art records with 72.77% ACC and 72.81% wa-F1 on IEMOCAP, and 65.44% ACC and 65.04% wa-F1 on MELD, respectively. The ablation results also indicate the effectiveness and rationality of each component in our approach.

The early detection of lung diseases, including cancer, is essential for improving patient outcomes. However, traditional diagnostic approaches and standard deep learning models often face challenges in effectively analyzing the complex spatial and temporal variations in medical imaging data, particularly in CT scans. To address these challenges, we propose DC-CAD, a novel hybrid framework that integrates Dilated Capsule Networks, Channel-wise Attention Mechanisms, and Distanced Long Short-Term Memory for precise and early diagnosis of lung diseases. DC-CAD is innovative in its ability to combine multi-scale feature extraction and temporal dynamic analysis, enabling the model to capture intricate spatial relationships and sequential changes in lung tissue. The model consists of three main contributions: (1) Dilated Capsule Networks for improved multi-scale context aggregation, which captures subtle textural variations, (2) a Channel-wise Attention Mechanism to focus on the most relevant regions of interest, minimizing the impact of irrelevant features, and (3) Distanced LSTM layers to model temporal dependencies across sequential CT scans, providing insights into disease progression. Through comprehensive experiments on the LC25000 dataset, DC-CAD achieves 99.52% accuracy, significantly outperforming baseline models such as standard Capsule Networks and Convolutional Neural Networks. The model also reduces the error rate to 0.48%, demonstrating substantial improvements in diagnostic performance, including increased accuracy, sensitivity, and specificity. These results establish DC-CAD as a powerful and reliable tool for automated lung disease diagnosis, with significant potential to enhance clinical workflow by reducing radiologists’ workload through its interpretability and efficiency. Moving forward, we plan to extend the model to handle multi-modal data and investigate advanced attention mechanisms to further improve diagnostic accuracy and generalizability.