Information Sciences最新文献

{"title":"Consensus-based iterative meta-pseudo-labeling for deep semi-supervised learning","authors":"David Aparco-Cardenas,&nbsp;Jancarlo F. Gomes,&nbsp;Alexandre X. Falcão,&nbsp;Pedro J. de Rezende","doi":"10.1016/j.ins.2024.121671","DOIUrl":"10.1016/j.ins.2024.121671","url":null,"abstract":"<div><div>A known issue that hinders the development of deep learning models is the need for accurate annotation of a large quantity of samples – a time-consuming, labor-intensive, and error-prone task. This limitation is particularly critical in areas where data annotation requires expert knowledge. Semi-supervised learning methods, such as pseudo-labeling, can alleviate the problem by capitalizing on both limited labeled and plentiful unlabeled data; nonetheless, state-of-the-art methods often require pre-trained encoders and validation sets to deliver effective solutions. Herein, we introduce a teacher-student-based iterative meta-pseudo-labeling approach, named consensus Deep Feature Annotation (<em>cons-DeepFA</em>), that enables the training of custom Convolutional Neural Networks (CNNs) from small quantities of labeled samples without reliance on pre-trained encoders and validation sets. cons-DeepFA explores <em>Feature Learning from Image Markers</em> (FLIM) to initialize the filters of a target CNN (student) from minimal data annotation – <em>i.e.</em>, user-drawn markers on discriminative regions of a few selected images per class. During each of a few iterations, the latent space of the student's last dense layer is non-linearly projected onto a two-dimensional space for downstream label propagation via an optimum-connectivity-based approach (teacher); afterward, the student is re-trained using pseudo-labeled samples selected by the proposed consensus mechanism, which jointly improves the latent space, its projection, and the student's generalization ability as iterations progress. This strategy was recently introduced with pre-trained encoders by selecting the most confident pseudo-labeled samples to re-train the student. While building on previous methods, cons-DeepFA presents two key contributions. It (i) incorporates FLIM to enable training a custom CNN from scratch with faster convergence, improving its generalization ability, and (ii) introduces a consensus-based procedure over multiple iterations that selects more accurately pseudo-labeled samples for re-training the CNN. Lastly, cons-DeepFA is evaluated on five challenging biological image datasets, demonstrating its effectiveness and competitiveness when compared to seven state-of-the-art methods from four semi-supervised learning paradigms.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"692 ","pages":"Article 121671"},"PeriodicalIF":8.1,"publicationDate":"2024-11-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142704064","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}

{"title":"Group search optimization-assisted deep reinforcement learning intelligence decision for virtual network mapping","authors":"Xiancui Xiao,&nbsp;Feng Yuan","doi":"10.1016/j.ins.2024.121664","DOIUrl":"10.1016/j.ins.2024.121664","url":null,"abstract":"<div><div>Virtual network mapping (VNM), as a key technology in network virtualization, has received widespread attention due to its ability to instantiate network services on infrastructure. However, existing VNM technologies have drawbacks, such as poor dynamic mapping processes, single search strategies, and low resource utilization. In this end, we propose a novel group search optimization-assisted deep reinforcement learning (DRL) intelligence decision for virtual network mapping, GSRL-VNM. In this algorithm, we first formalize the deep reinforcement learning model of VNM and describe the dynamic characteristics of VNM process. Then, in order to effectively reduce resource fragmentation and improve the mapping success rate in VNM process, group search optimization (GSO), a swarm intelligent optimization algorithm with excellent global search ability, is utilized to assist deep reinforcement learning intelligent decision-making by improving convergence speed and optimal value. The simulation results show that the proposed GSRL-VNM algorithm outperforms the existing baseline algorithms in terms of acceptance rate, link pressure, long-term average cost, and average revenue.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"693 ","pages":"Article 121664"},"PeriodicalIF":8.1,"publicationDate":"2024-11-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142706184","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}

{"title":"Restoration after deterioration in interdependent infrastructure networks: A two-stage hybrid method with minimum network performance loss","authors":"Yulong Li ,&nbsp;Han Su ,&nbsp;Baisong Yang ,&nbsp;Jie Lin ,&nbsp;Yinghua Shen ,&nbsp;Guobin Wu","doi":"10.1016/j.ins.2024.121655","DOIUrl":"10.1016/j.ins.2024.121655","url":null,"abstract":"<div><div>Restoration of interdependent infrastructure networks (IINs) relies on the information from deterioration, which is of great significance because IINs support the normal functioning of social productivity and life. However, existing research has not fully addressed the restoration after deterioration in IINs, which is not conducive to the timely elimination of the adverse effects of infrastructure deterioration. First, a unified model for IINs is innovatively devised by considering both functional and operational interdependencies between infrastructures. Second, a two-stage hybrid method that generates the optimal restoration strategies after deterioration in IINs is proposed. Specifically, in the first stage, a hidden Markov chain model for deterioration prediction is constructed, which is solved by the Expectation-Maximization (EM) algorithm. In the second stage, an objective programming model with minimum network performance loss for restoration optimization is developed, and the optimal strategy is obtained by the ant colony algorithm. Finally, a real-world case is used to validate the feasibility and effectiveness of the proposed method. The results show that this method is efficient and effective in finding optimal restoration strategy after deterioration in IINs. We also investigate the effects of initial restoration time and restoration resource grouping, which provide helpful decision guidance for real cases.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"693 ","pages":"Article 121655"},"PeriodicalIF":8.1,"publicationDate":"2024-11-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142706185","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}

{"title":"Improving stochastic models by smart denoising and latent representation optimization","authors":"Jakob Jelenčič ,&nbsp;M. Besher Massri ,&nbsp;Ljupčo Todorovski ,&nbsp;Marko Grobelnik ,&nbsp;Dunja Mladenić","doi":"10.1016/j.ins.2024.121672","DOIUrl":"10.1016/j.ins.2024.121672","url":null,"abstract":"<div><div>This paper introduces an innovative deep learning-based optimization method specifically designed for data derived from stochastic processes. Addressing the prevalent issue of rapid overfitting in real-world scenarios with limited historical data, our approach focuses on denoising optimization. The method effectively balances the simultaneous optimization of latent data representation and target variables, leading to enhanced model performance. We rigorously test our approach using five diverse real-world datasets. Our study is structured into three parts: an ablation study to validate the individual components of our method, a statistical analysis using the Wilcoxon rank-sum test to confirm the superiority of our method against five research hypotheses, and a detailed exploration of parameter visualization and fine-tuning. The comprehensive evaluation demonstrates that our method not only outperforms existing techniques but also significantly contributes to the advancement of deep learning models for stochastic processes. The findings underscore the potential of our method as a robust solution to the challenges in modeling stochastic processes with deep learning, offering new avenues for efficient and accurate predictions.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"692 ","pages":"Article 121672"},"PeriodicalIF":8.1,"publicationDate":"2024-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142703996","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}

{"title":"Preserving privacy in association rule mining using multi-threshold particle swarm optimization","authors":"Shahad Aljehani ,&nbsp;Youseef Alotaibi","doi":"10.1016/j.ins.2024.121673","DOIUrl":"10.1016/j.ins.2024.121673","url":null,"abstract":"<div><div>Healthcare data has become a powerful resource for generating insights that drive medical research. Association Rule Mining (ARM) techniques are widely used to identify relationships among diseases, treatments, and symptoms. However, sensitive information is often exposed, creating significant privacy challenges, particularly when data is integrated from multiple sources. Although Privacy-Preserving Association Rule Mining (PPARM) methods have been developed to address these issues, most rely on a single, predefined Minimum Support Threshold (MST) that is inflexible in adapting to diverse rule patterns. In this study, a Multi-Threshold Particle Swarm Optimization for Association Rule Mining (MPSO4ARM) model is introduced, integrating the Apriori and Particle Swarm Optimization (PSO) algorithms to perform data mining while protecting sensitive rules. A novel approach is employed by the proposed model to dynamically adjust the MST, allowing for more adaptive and effective privacy preservation. The MPSO4ARM model adjusts the MST on-the-fly based on rule length, improving its ability to safeguard sensitive data across various datasets. The proposed model was evaluated on the Chess, Mushroom, Retail, and Heart Disease datasets. The experimental results showed that the MPSO4ARM model outperforms traditional Apriori and conventional PSO algorithms, achieving higher fitness values and reducing side effects such as Hiding Failure (HF) and Missing Cost (MC), particularly in the Heart Disease and Mushroom datasets. Although the dynamic MST function introduces a moderate increase in computational runtime compared to Apriori and conventional PSO, this trade-off between execution time and enhanced privacy protection is considered acceptable, given the model's substantial improvements in data utility and rule sanitization.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"692 ","pages":"Article 121673"},"PeriodicalIF":8.1,"publicationDate":"2024-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142704000","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}

{"title":"On the exploitation of control knowledge for enhancing automated planning","authors":"Xu Lu ,&nbsp;Bin Yu ,&nbsp;Cong Tian ,&nbsp;Chu Chen ,&nbsp;Zhenhua Duan","doi":"10.1016/j.ins.2024.121666","DOIUrl":"10.1016/j.ins.2024.121666","url":null,"abstract":"<div><div>Heuristic search provides an efficient way to automatically explore the state space in planning, while Control Knowledge (CK) also has the potential to significantly increase the performance of planners. Currently, most of the state-of-the-art planners primarily rely on sophisticated heuristic mechanisms. However, these planners fail to scale up and to provide (high-quality) solutions in a range of problems.</div><div>The objective of this paper is to incorporate CK with heuristic search in order to leverage the advantages of both, thus leading planners to achieve much higher efficiency. To achieve this, we introduce a novel CK which is specified by a variant of Linear Temporal Logic (LTL), referred to as LTL<span><math><msub><mrow></mrow><mrow><mi>P</mi></mrow></msub></math></span>. We propose an encoding methodology that translates LTL<span><math><msub><mrow></mrow><mrow><mi>P</mi></mrow></msub></math></span> into standard planning models. Consequently, we can directly use existing heuristic planners to solve the augmented problem, and avoid tailoring the planners in order to deal with CK implicitly. The novelty of this approach lies in that we define a useful CK LTL<span><math><msub><mrow></mrow><mrow><mi>P</mi></mrow></msub></math></span> with a concise encoding methodology, that can significantly improve the efficiency of heuristic search. In this paper, the encoding process is formally presented, and theoretical results on the complexity and soundness of the encoding are strictly proved. We find that appropriate CK is a good complement to heuristic search, and is capable of making hard problems easy to solve. Experiments demonstrate that our approach shows highly competitive results versus heuristic search and other CK-based techniques on many intractable benchmark problems, benefiting in improving the coverage and quality of plans.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"693 ","pages":"Article 121666"},"PeriodicalIF":8.1,"publicationDate":"2024-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142706186","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}

{"title":"An insightful data-driven crowd simulation model based on rough sets","authors":"Tomasz Hachaj,&nbsp;Jarosław Wąs","doi":"10.1016/j.ins.2024.121670","DOIUrl":"10.1016/j.ins.2024.121670","url":null,"abstract":"<div><div>Data-driven crowd simulation with insightful principles is an open, real-world, and challenging task. The issues involved in modeling crowd movement so that agents' decision-making processes can be interpreted provide opportunities to learn about the mechanisms of crowd formation and dispersion and how groups cope with overcoming obstacles. In this article, we propose a novel agent-based simulation algorithm to infer practical knowledge of a problem from the real world by modeling the domain knowledge available to an agent using rough sets. As far as we know, the method proposed in our work is the first approach that integrates a well-established agent-based simulation model of social forces, an insightful knowledge representation using rough sets, and Bayes probability inference that models the stochastic nature of motion. Our approach has been tested on real datasets representing crowds traversing bottlenecks of varying widths. We also conducted a test on numerous artificial datasets involving 1,000 agents. We obtained satisfactory results that confirm the effectiveness of the proposed method. The dataset and source codes are available for download so our experiments can be reproduced.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"692 ","pages":"Article 121670"},"PeriodicalIF":8.1,"publicationDate":"2024-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142704058","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}

{"title":"Enabling multi-step forecasting with structured state space learning module","authors":"Shaoqi Wang,&nbsp;Chunjie Yang","doi":"10.1016/j.ins.2024.121669","DOIUrl":"10.1016/j.ins.2024.121669","url":null,"abstract":"<div><div>Data-driven soft sensor incorporated with the model predictive control (MPC) algorithms facilitating product quality and cost control is of imperative importance in industrial processes. However, the widely used one-step forecasting method can not incorporate with MPC and therefore restricts the practical usage of soft sensor. Multi-step forecasting introduces long-term dependencies problems yet has not been effectively resolved within traditional model structure. To address this problem, this paper proposes the deep learning network architecture named Extended State Space Learning Module (ESSLM). ESSLM extends the nonlinear mapping architecture of deep learning based on state space and retains state transfer matrices to characterize the dynamics of the system. ESSLM distinguishes itself from explicit network architectures such as gated RNNs by addressing the long-term dependencies problems through an implicit initialization method, and the MLP and RNN algorithms can be regarded as the manifestation of ESSLM in special cases. ESSLM characterizes the latent space as the coefficients of the orthogonal basis functions so that the input data can be encoded into a high-dimensional feature space with minimal information loss which efficiently achieves multi-step forecasting and give greater utility and practical significance.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"692 ","pages":"Article 121669"},"PeriodicalIF":8.1,"publicationDate":"2024-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142704076","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}

{"title":"Improving diversity and invariance for single domain generalization","authors":"Zhen Zhang ,&nbsp;Shuai Yang ,&nbsp;Qianlong Dang ,&nbsp;Tingting Jiang ,&nbsp;Qian Liu ,&nbsp;Chao Wang ,&nbsp;Lichuan Gu","doi":"10.1016/j.ins.2024.121656","DOIUrl":"10.1016/j.ins.2024.121656","url":null,"abstract":"<div><div>Single domain generalization aims to train a model that can generalize well to multiple unseen target domains by leveraging the knowledge in a related source domain. Recent methods focus on synthesizing domains with new styles to improve the diversity of training data. However, mainstream methods rely heavily on an additional generative model when generating augmented data, which increases optimization difficulties and is not conducive to generating diverse style data. Moreover, these methods do not sufficiently capture the consistency between the generated and original data when learning feature representations. To address these issues, we propose a novel single domain generalization method, namely DAI, which improves <strong>D</strong>iversity <strong>A</strong>nd <strong>I</strong>nvariance simultaneously to boost the generalization capability of the model. Specifically, DAI consists of a style diversity module and a representation learning module optimized in an adversarial learning manner. The style diversity module uses a generative model, nAdaIN, to synthesize the data with significant style shifts. The representation learning module performs object-aware contrastive learning to capture the invariance between the generated and original data. Furthermore, DAI progressively synthesizes multiple novel domains to increase the style diversity of generated data. Experimental results on three benchmarks show the superiority of our method against domain shifts.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"692 ","pages":"Article 121656"},"PeriodicalIF":8.1,"publicationDate":"2024-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142704057","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}

{"title":"A novel decision-making agent-based multi-objective automobile insurance pricing algorithm with insurers and customers satisfaction","authors":"Tao Ma ,&nbsp;Li Guang Xie ,&nbsp;Hong Zhao ,&nbsp;Fang Yang ,&nbsp;Chunsheng Liu ,&nbsp;Jing Liu","doi":"10.1016/j.ins.2024.121665","DOIUrl":"10.1016/j.ins.2024.121665","url":null,"abstract":"<div><div>Automobile insurance is essential for customers to obtain an appropriate automobile insurance policy and plays a significant role in the insurance industry. However, pricing a reasonable automobile insurance policy for different customers requires considering several indicators, such as the number of historical accidents, the frequency of long-distance driving, and the mileage driven per month. Therefore, how to make a reasonable automobile insurance pricing policy according to the needs of different customers is an urgent issue. Existing pricing methods often face the challenges of inefficiencies and mispricing, making it difficult to satisfy both insurers and customers. To better address this problem, this paper proposes a decision-making agent-based multi-objective automobile insurance pricing (DMA-MoAIP) algorithm. The DMA-MoAIP algorithm provides a diverse set of pricing strategies that meet various requirements, and considers the satisfaction of both insurers and customers. Firstly, a decision-making agent (DMA) framework is proposed for insurers, which provides an accurate assessment of the risk level for each customer. Secondly, a data-driven scoring (DDS) mechanism is established to better measure the satisfaction of insurers and customers and form a score that corresponds to their satisfaction. Thirdly, a novel multi-objective particle swarm optimization (MoPSO) algorithm is designed to search for effective and diverse solutions in dealing with automobile insurance pricing problem. Experimental results show that DMA-MoAIP outperforms five state-of-the-art multi-objective algorithms (including NSGA-II and so on) in terms of convergence and solution diversity. Specifically, the solution diversity of DMA- MoAIP in automotive pricing has increased by an average of 17%, which can provide more pricing options to meet different customer needs. In practical applications, DMA-MoAIP provides three distinct pricing strategies: insurer-oriented, customer-oriented, and compromise pricing. These strategies underscore the importance of considering relevant driving behavior metrics, rendering them valuable for real-life applications.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"693 ","pages":"Article 121665"},"PeriodicalIF":8.1,"publicationDate":"2024-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142706182","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}