Applied Soft Computing最新文献

筛选
英文 中文
Multi-Dueling framework for multi-agent reinforcement learning 多智能体强化学习的多决斗框架
IF 7.2 1区 计算机科学
Applied Soft Computing Pub Date : 2025-06-21 DOI: 10.1016/j.asoc.2025.113464
Baochang Ren , Mingjie Cai , Bin Yu
{"title":"Multi-Dueling framework for multi-agent reinforcement learning","authors":"Baochang Ren ,&nbsp;Mingjie Cai ,&nbsp;Bin Yu","doi":"10.1016/j.asoc.2025.113464","DOIUrl":"10.1016/j.asoc.2025.113464","url":null,"abstract":"<div><div>In real-world tasks, multiple agents often need to coordinate with one another due to their individual private observations and restricted communication abilities. A representative research direction is the deep multi-agent reinforcement learning value decomposition, which decomposes the global shared joint action value function <span><math><mrow><msub><mrow><mi>Q</mi></mrow><mrow><mi>t</mi><mi>o</mi><mi>t</mi></mrow></msub><mrow><mo>(</mo><mi>τ</mi><mo>,</mo><mi>u</mi><mo>)</mo></mrow></mrow></math></span> into their respective action value functions <span><math><mrow><msub><mrow><mi>Q</mi></mrow><mrow><mi>i</mi></mrow></msub><mrow><mo>(</mo><msub><mrow><mi>τ</mi></mrow><mrow><mi>i</mi></mrow></msub><mo>,</mo><msub><mrow><mi>u</mi></mrow><mrow><mi>i</mi></mrow></msub><mo>)</mo></mrow></mrow></math></span> to guide the behavior of individuals. They all follow the IGM (Individual-Global-Max) principle, obeying the addable assumption and the monotonic assumption to support effective local decision making. However, to achieve scalability, existing MARL algorithms often compromise either the expressive power of their value function representations or the consistency of the IGM principles. This compromise can potentially result in instability or poor performance when dealing with complex tasks. In this paper, we introduce a novel algorithm called MDF—a Multi-Dueling Framework for Multi-Agent Reinforcement Learning. We innovatively propose the V-IGM constraint principle and correct the incomplete expression of the constant term <span><math><mrow><mi>c</mi><mrow><mo>(</mo><mi>τ</mi><mo>)</mo></mrow></mrow></math></span> of the Qatten algorithm to further refine the decomposition of the joint action value function <span><math><mrow><msub><mrow><mi>Q</mi></mrow><mrow><mi>t</mi><mi>o</mi><mi>t</mi></mrow></msub><mrow><mo>(</mo><mi>τ</mi><mo>,</mo><mi>u</mi><mo>)</mo></mrow></mrow></math></span>. The MDF algorithm innovatively utilizes the Dueling Network architecture for decomposing the joint action value function <span><math><mrow><msub><mrow><mi>Q</mi></mrow><mrow><mi>t</mi><mi>o</mi><mi>t</mi></mrow></msub><mrow><mo>(</mo><mi>τ</mi><mo>,</mo><mi>u</mi><mo>)</mo></mrow></mrow></math></span>. Additionally, it incorporates the multi-attention mechanism to achieve an even more refined decomposition of <span><math><mrow><msub><mrow><mi>Q</mi></mrow><mrow><mi>t</mi><mi>o</mi><mi>t</mi></mrow></msub><mrow><mo>(</mo><mi>τ</mi><mo>,</mo><mi>u</mi><mo>)</mo></mrow></mrow></math></span>. Experiments show that MDF algorithm outperforms the most advanced MARL algorithm in StarCraft<span><math><mi>Π</mi></math></span> maps (e.g. 3 m, 8 m, 2m-vs-1z, 2m-vs-1sc, 2s3z, 3s-vs-4z, 3s-vs-5z, 3s5z, 1c3s5z, bane-vs-bane).</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"181 ","pages":"Article 113464"},"PeriodicalIF":7.2,"publicationDate":"2025-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144338740","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
A neural network based on back-propagation and cooperative co-evolution 基于反向传播和协同进化的神经网络
IF 7.2 1区 计算机科学
Applied Soft Computing Pub Date : 2025-06-21 DOI: 10.1016/j.asoc.2025.113453
Yuelin Gao , Yuming Zhang , Xiaofeng Xie
{"title":"A neural network based on back-propagation and cooperative co-evolution","authors":"Yuelin Gao ,&nbsp;Yuming Zhang ,&nbsp;Xiaofeng Xie","doi":"10.1016/j.asoc.2025.113453","DOIUrl":"10.1016/j.asoc.2025.113453","url":null,"abstract":"<div><div>Deep neural networks (DNNs) have a powerful feature extraction capability, which allows them to be employed in various fields. However, as the number of layers and neurons in the network increases, the search space for parameter learning becomes complex. Currently, the most commonly used parameter training method is backpropagation (BP) based on gradient descent, but this method is sensitive to the initialization of the parameters and tends to get stuck in local optima in a complex search space. Therefore, a new training method for DNNs has been proposed that combines cooperative co-evolution (CC) with BP-based gradient descent, called BPCC. In the BPCC method, BP performs multiple training periods intermittently, and the CC algorithm is executed when the difference between the current loss function value and the previous loss function value is less than a given threshold (called a condition met). We found that the algorithm easily enters into CC iterations, which reduces the computational effectiveness of the algorithm. A tolerance parameter is designed to curb this phenomenon, and the CC is executed when the cumulative number of times the condition is met reaches the given value of the tolerance parameter, and the improved gray wolf optimizer (GWO) algorithm is used as the solver for the CC. In addition, in the CC iteration stage, the Chebyshev chaotic map series based on the current optimal point is used to initialize the population of GWO to ensure the diversity of the initial population. Experimental comparisons are made with modern network training methods in 7 network models, and the experimental results show that the improved algorithm in this study is competitive.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"181 ","pages":"Article 113453"},"PeriodicalIF":7.2,"publicationDate":"2025-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144366746","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Lightweight decision-making decisive feature enhancement network for medical image analysis 用于医学图像分析的轻量级决策决定性特征增强网络
IF 7.2 1区 计算机科学
Applied Soft Computing Pub Date : 2025-06-21 DOI: 10.1016/j.asoc.2025.113518
Xiangyang Ren , Boyang Jiao , Jianbo Gao , Yazheng Chen , Na Xiao , Ying Bi , Gangqiong Liu
{"title":"Lightweight decision-making decisive feature enhancement network for medical image analysis","authors":"Xiangyang Ren ,&nbsp;Boyang Jiao ,&nbsp;Jianbo Gao ,&nbsp;Yazheng Chen ,&nbsp;Na Xiao ,&nbsp;Ying Bi ,&nbsp;Gangqiong Liu","doi":"10.1016/j.asoc.2025.113518","DOIUrl":"10.1016/j.asoc.2025.113518","url":null,"abstract":"<div><div>Medical image segmentation is crucial in diagnosing and treating various diseases. Most existing medical segmentation methods often overlook the importance of selecting decision features, resulting in the extraction of redundant target features, which often leads to a large number of model parameters and poor deployability. Therefore, to reduce the parameter count of medical image segmentation models and improve their deployability, we propose a two-phase detection network based on enhancing decision-making decisive (DMD) features, termed the Decision-Making Decisive Feature Enhancement Network (DDFE-Net). The core idea of DDFE-net is to reduce the number of parameters required for model fitting and redundant target features by screening and enhancing the features that are important for decision-making. Specifically, in the DDFE-net, we first propose a decision network (DE-net) for initially screening and extracting DMD features through dense multi-level feature fusion and deep supervision. The DMD features of medical targets are effectively extracted through dense multi-level feature extraction and fusion. Subsequently, we introduced a DMD feature enhancement network (DEE-net) into the DDFE network to enhance the feature representation of medical targets. The DEE-net integrates DMD features of different scales and levels in the DE-net by performing secondary encoding and decoding on the extracted DMD features, thereby achieving DMD feature enhancement and further eliminating redundant features, reducing the number of model parameters, and improving the network's feature expression ability. Extensive experimental results on several medical segmentation benchmark datasets, prove that the proposed DDFE-net outperforms other state-of-the-art (SOTA) methods by 8 % in accuracy and achieves a 49 % reduction in model size, greatly improving the deployability of medical image segmentation methods.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"181 ","pages":"Article 113518"},"PeriodicalIF":7.2,"publicationDate":"2025-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144366748","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Optimizing multi-focus image fusion through convolutional attention vision transformers and spatial consistency models 基于卷积注意视觉变换和空间一致性模型的多焦点图像融合优化
IF 7.2 1区 计算机科学
Applied Soft Computing Pub Date : 2025-06-20 DOI: 10.1016/j.asoc.2025.113507
Shengchuan Jiang , Shanchuan Yu
{"title":"Optimizing multi-focus image fusion through convolutional attention vision transformers and spatial consistency models","authors":"Shengchuan Jiang ,&nbsp;Shanchuan Yu","doi":"10.1016/j.asoc.2025.113507","DOIUrl":"10.1016/j.asoc.2025.113507","url":null,"abstract":"<div><div>Multi-Focus Image Fusion (MFIF) aims to integrate all pixels in the images to avoid defocused pixels simultaneously. The removal of defocused pixels in the image is more challenging because the traditional approach makes it difficult to accurately detect defocused regions in the image. In this paper, the Convolutional Attention based Vision Transformer based Iterative Multi-Scale Fusion Network (CAViT-IMSFN) model is proposed for MFIF. The collected input images are fed into preprocessing approaches like normalization, data imputation, and augmentation for improving the generalization ability of the model and also for reducing overfitting issues. The convolutional-based model named MobileNetV2 is used to extract local features and the AViT model is introduced to extract global features present in the preprocessed images. The spatial attention model is used in the integration stage to integrate both local and global features that preserve spatial consistency. For enhancing image quality and spatial consistency, the iterative refinement model is implemented according to the feedback mechanism that helps to update the fused output iteratively. The gradient boosting optimization algorithm is used for weight adjustment and the multi-scale fusion model is applied for the identification of focused and defocused portions in the images. This proposed model improves network adaptability by capturing multi-scale features and also has the ability to handle various levels of detail and complexity. The experimental evaluation is performed in terms of using diverse performance evaluation measures and quantitative analyses. The proposed model attained performances of 1.435 from Normalized Mutual Information (NMI) and 0.78 s from computational time. The result showed that the proposed model achieved superior outcomes rather than other MFIF-related existing approaches.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"181 ","pages":"Article 113507"},"PeriodicalIF":7.2,"publicationDate":"2025-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144502722","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Adaptive temporal diffusion-based reconstruction model for industrial dynamic uncertain process monitoring 基于自适应时间扩散的工业动态不确定过程监控重建模型
IF 7.2 1区 计算机科学
Applied Soft Computing Pub Date : 2025-06-20 DOI: 10.1016/j.asoc.2025.113407
Jiawei Yin , Jianbo Yu , Qingchao Jiang , Xuefeng Yan
{"title":"Adaptive temporal diffusion-based reconstruction model for industrial dynamic uncertain process monitoring","authors":"Jiawei Yin ,&nbsp;Jianbo Yu ,&nbsp;Qingchao Jiang ,&nbsp;Xuefeng Yan","doi":"10.1016/j.asoc.2025.113407","DOIUrl":"10.1016/j.asoc.2025.113407","url":null,"abstract":"<div><div>A key property of industrial processes is that they are often related to the dynamic uncertainty behaviors of the measurement data (e.g., sensor performance degradation or environmental changes), which poses significant challenges for traditional uncertainty-based monitoring research that typically assumes the measurement data exhibits invariant uncertainty. This study addresses this challenge by enriching the corrupted state of the data. We propose a novel diffusion-based method called the Dynamic Uncertainty Process Monitoring (DUPM) method. DUPM consists of a temporal diffusion convolutional network (TDCN) module and an adaptive diffusion reconstruction (ADR) module. First, in TDCN module, the diffusion process enhances the pattern coverage by gradually corrupting the original data, enabling the model to cover data under different uncertainties. Then an unsupervised backbone is designed to extract the latent temporal features of the input data and remove the noise in generation process, in which a nonlinear autoencoder is equipped with a one-dimensional convolution operation. The monitoring threshold is determined based on the reconstruction error after the diffusion process and generation process. Finally, the ADR module determines the number of steps to add noise in the online stage by calculating the similarity between the online data and the historical diffusion states. In this way, the reconstruction error can be used as a monitoring score. Experiments conducted on numerical simulations and the real-world MetroPT-3 dataset show that DUPM reduces the false alarm rate by at least 3% compared to the comparison method while maintaining fault detection rates. The verification indicates that the proposed method has potential in monitoring dynamic uncertain industrial processes.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"181 ","pages":"Article 113407"},"PeriodicalIF":7.2,"publicationDate":"2025-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144366749","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Enhancing knowledge distillation via genetic recombination 通过基因重组增强知识精馏
IF 7.2 1区 计算机科学
Applied Soft Computing Pub Date : 2025-06-20 DOI: 10.1016/j.asoc.2025.113414
Yangjie Cao , Chuanjin Zhou , Minglin Liu , Weiqi Luo , Xiangyang Luo
{"title":"Enhancing knowledge distillation via genetic recombination","authors":"Yangjie Cao ,&nbsp;Chuanjin Zhou ,&nbsp;Minglin Liu ,&nbsp;Weiqi Luo ,&nbsp;Xiangyang Luo","doi":"10.1016/j.asoc.2025.113414","DOIUrl":"10.1016/j.asoc.2025.113414","url":null,"abstract":"<div><div>Diverging from conventional knowledge distillation methods that solely emphasize improving the utilization of the teacher’s knowledge, this paper explores the generation of stronger student models within available knowledge. We first conceptualize the knowledge distillation process as a genetic evolution model. The student model is regarded as an independent individual, with its parameters representing the genes of that individual. These genes are partitioned into several alleles according to the architecture of the student model. Following that, we propose a universal strategy to enhance existing knowledge distillation methods by introducing genetic recombination. Prior to distillation, we initialize two independent identically distributed student models with different random seeds to obtain the first generation of genes. With each epoch of distillation, these genes evolve into the next generation. At specific generations, we randomly select one exchangeable allele from each of the two students for exchange. Our focus lies in determining the alleles to exchange and their corresponding exchange frequency (i.e., crossing-over value). This approach provides more choices and possibilities for subsequent evolution. Extensive experiments confirm the effectiveness of the strategy, demonstrating improvements across 12 distillation methods and 17 teacher–student combinations.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"181 ","pages":"Article 113414"},"PeriodicalIF":7.2,"publicationDate":"2025-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144335810","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Multi-agent cooperative multi-network group framework for energy-efficient distributed fuzzy flexible job shop scheduling problem 节能分布式模糊柔性作业车间调度问题的多智能体协同多网络群框架
IF 7.2 1区 计算机科学
Applied Soft Computing Pub Date : 2025-06-19 DOI: 10.1016/j.asoc.2025.113474
Zi-Qi Zhang , Xiao-Wei Li , Bin Qian , Huai-Ping Jin , Rong Hu , Jian-Bo Yang
{"title":"Multi-agent cooperative multi-network group framework for energy-efficient distributed fuzzy flexible job shop scheduling problem","authors":"Zi-Qi Zhang ,&nbsp;Xiao-Wei Li ,&nbsp;Bin Qian ,&nbsp;Huai-Ping Jin ,&nbsp;Rong Hu ,&nbsp;Jian-Bo Yang","doi":"10.1016/j.asoc.2025.113474","DOIUrl":"10.1016/j.asoc.2025.113474","url":null,"abstract":"<div><div>The increasing integration of industrial intelligence and the Industrial Internet of Things (IIoT) has promoted distributed flexible manufacturing (DFM) as a fundamental component of smart manufacturing systems. However, the rising complexity in dynamic demands, production uncertainties, and the urgent need for energy efficiency pose significant challenges. To address these challenges, this study investigates the energy-efficient distributed fuzzy flexible job shop scheduling problem (EE-DFFJSP), which aims to minimize both makespan and total energy consumption (TEC) in DFM environments. To tackle fuzzy uncertainties and complex coupling characteristics inherent in EE-DFFJSP, a multi-agent cooperative multi-network group (MACMNG) framework is proposed. First, a mixed-integer linear programming (MILP) model for EE-DFFJSP is formulated, followed by an analysis of the problem’s properties. A triple Markov decision process formulation adapted to the problem's characteristics is designed, enabling problem decoupling and multi-agent decision-making through specific state representations and reward functions. Next, an innovative multi-network group framework is devised, and coupled decisions are effectively handled via interaction and collaboration among independent subnets. Based on problem decomposition method, EE-DFFJSP is decomposed into a set of subproblems represented by subnets within the network group. These subnets cooperate by sharing experience and knowledge through a domain parameter transfer strategy (DPTS) to enable efficient training. Finally, MACMNG employs a multi-objective DQN (MO-DQN) integrated with a dynamic weighting mechanism, enabling subnets to effectively balance between makespan and TEC during cooperative decision-making and network parameter updating. Experimental results show that MACMNG achieves superior performance compared with three priority dispatch rules (PDRs) across various scenarios. The MACMNG outperforms seven state-of-the-art multi-objective algorithms in terms of different metrics across 69 benchmark instances. This study contributes an efficient learning-driven and multi-agent collaborative promising paradigm for the energy-efficient scheduling in DFM, providing practical insights for advancing smart manufacturing in IIoT architectures.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"181 ","pages":"Article 113474"},"PeriodicalIF":7.2,"publicationDate":"2025-06-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144470732","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
An integrated decision-making framework for evaluating Industry 5.0 and Circular Economy in supply chain management using Z-numbers 供应链管理中工业5.0和循环经济评价的综合决策框架
IF 7.2 1区 计算机科学
Applied Soft Computing Pub Date : 2025-06-19 DOI: 10.1016/j.asoc.2025.113504
Seyyed Jalaladdin Hosseini Dehshiri
{"title":"An integrated decision-making framework for evaluating Industry 5.0 and Circular Economy in supply chain management using Z-numbers","authors":"Seyyed Jalaladdin Hosseini Dehshiri","doi":"10.1016/j.asoc.2025.113504","DOIUrl":"10.1016/j.asoc.2025.113504","url":null,"abstract":"<div><div>Due to competitive pressures and awareness of environmental issues, companies focus on improving resilience and sustainability in the Supply Chain (SC). Industry 5.0 (I5.0) and artificial intelligence are two examples of new technologies that can improve SCs' resilience, efficiency, and transparency. Additionally, the Circular Economy (CE) supports sustainability by promoting reuse, recycling, and reduction of waste and resource consumption. This research proposes integrating I5.0 and CE in SC to achieve sustainability. A decision approach using Z-numbers is developed to evaluate the solutions. A novel integrated framework including the Simplified Best-Worst Method (SBWM), and Combined Compromise Solution (CoCoSo) approaches, is extended using Z-numbers to evaluate the implementation solutions of I5.0 and CE in SC. This approach offers flexible answers and reliable findings based on different decision-making situations. Also, comparative analysis based on different techniques and sensitivity analysis are investigated. Paying attention to appropriate investment cost, suitable investment risk, and green factors and the growth of environmentally affable procedures were the most significant sub-criteria with the significance of 0.232, 0.150, and 0.139, respectively. The results indicated that the solutions for developing digital infrastructure and suitable information systems, providing financial resources and development and attraction of investment, and creating the information system for tracking products in the circular SC are the most appropriate implementation solutions in the Z-CoCoSo method with scores of 4.703, 4.335, and 3.864, respectively. The findings of comparative analysis and examination of various scenarios also confirmed the priority of the solutions. Digital infrastructure and information systems enhance coordination and speed in SCs. Investing in advanced technologies, upgrading equipment, and attracting investors can mitigate financial risks. Additionally, using product tracking systems boosts transparency, supports sustainability, and ensures compliance with environmental regulations.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"181 ","pages":"Article 113504"},"PeriodicalIF":7.2,"publicationDate":"2025-06-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144513763","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
SANTM: A Sparse Access Neural Turing Machine with local multi-head self-attention for long-term memorization 基于局部多头自注意的长期记忆稀疏访问神经图灵机
IF 7.2 1区 计算机科学
Applied Soft Computing Pub Date : 2025-06-19 DOI: 10.1016/j.asoc.2025.113448
Dongjing Shan , Jing Zhu , Yamei Luo
{"title":"SANTM: A Sparse Access Neural Turing Machine with local multi-head self-attention for long-term memorization","authors":"Dongjing Shan ,&nbsp;Jing Zhu ,&nbsp;Yamei Luo","doi":"10.1016/j.asoc.2025.113448","DOIUrl":"10.1016/j.asoc.2025.113448","url":null,"abstract":"<div><div>In this paper, we propose a Sparse Access Neural Turing Machine (SANTM) to address long-term memorization challenges in sequence learning. The SANTM integrates a three-level neural controller with external memory: (1) a bottom layer for segmenting inputs into variable-length chunks, (2) a middle layer for short-term memory integration, and (3) a top layer that selectively accesses external memory via a locality-biased multi-head self-attention mechanism based on ChebNet spectral graph convolution. A sparse mask, trained through an <span><math><msub><mrow><mi>ℓ</mi></mrow><mrow><mn>0</mn></mrow></msub></math></span>-constrained optimization scheme, reduces memory access rates while enabling pre-fetching. Theoretical analysis derives an optimal access rate under idealized conditions. Experiments on sequential image classification (MNIST, CIFAR10), text classification, speaker discrimination, and language modeling (WikiText-103, enwik8) demonstrate SANTM’s superiority over state-of-the-art sequential models. Key results include 95.7% accuracy on permuted MNIST (vs. NTM’s 94.0%), 85.4% on TC-Speech (vs. 79.6% for NTM), and 24.2 perplexity on WikiText-103 (vs. Transformer-XL’s 27.0). The sparse mask reduces FLOPs by 37%–54% compared to traditional NTMs, validating its efficiency.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"181 ","pages":"Article 113448"},"PeriodicalIF":7.2,"publicationDate":"2025-06-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144330812","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Opinions-integrated consensus models for large-scale group decision-making in emergency medical capability evaluation 应急医疗能力评估中大规模群体决策的意见整合共识模型
IF 7.2 1区 计算机科学
Applied Soft Computing Pub Date : 2025-06-18 DOI: 10.1016/j.asoc.2025.113494
Xiaoting Cheng , Kai Zhang , Zeshui Xu , Xunjie Gou
{"title":"Opinions-integrated consensus models for large-scale group decision-making in emergency medical capability evaluation","authors":"Xiaoting Cheng ,&nbsp;Kai Zhang ,&nbsp;Zeshui Xu ,&nbsp;Xunjie Gou","doi":"10.1016/j.asoc.2025.113494","DOIUrl":"10.1016/j.asoc.2025.113494","url":null,"abstract":"<div><div>Emergency medical capability is critical for community resilience and emergency response. However, existing evaluation methods mainly rely on expert insights while ignoring public perspectives. To bridge this gap, two opinions-integrated consensus models for large-scale group decision-making (LSGDM) are proposed. First, public opinions are analyzed using fuzzy-set Qualitative Comparative Analysis to determine criteria weights. An importance slider and programming model are introduced to quantify the relative importance of public opinions. A backtracking identification method is introduced to adjust expert insights and facilitate consensus. Based on these, a comprehensive consensus model and a professional consensus model are developed. Simulation and sensitivity analysis demonstrate the effectiveness of both models in consensus reaching. Overall, the professional consensus model performs better due to its stricter judgment mechanism. Additionally, the performance of both models is sensitive to parameter settings. Accordingly, the adaptability of both models is further discussed in terms of public participation and acceptance, evaluation timeliness, and expert heterogeneity. This study provides a systematic approach to integrating public opinions and expert insights in LSGDM, enhancing the credibility and applicability of evaluation results.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"181 ","pages":"Article 113494"},"PeriodicalIF":7.2,"publicationDate":"2025-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144470730","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
相关产品
×
本文献相关产品
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信