International Journal of Approximate Reasoning最新文献

{"title":"Three-way conflict analysis: From single-level to multi-level preferences","authors":"Guangming Lang ,&nbsp;Haojun Liu ,&nbsp;Mengjun Hu","doi":"10.1016/j.ijar.2025.109621","DOIUrl":"10.1016/j.ijar.2025.109621","url":null,"abstract":"<div><div>In three-way conflict analysis, a key challenge lies in faithfully capturing agents’ attitudes towards multiple issues within complex conflict situations. Preference-based conflict situations, which characterize comparisons through ordered issue pairs, offer a structured alternative to traditional rating scales by emphasizing relational judgments. However, existing single-level preference frameworks are limited in their ability to capture variations in preference strengths, confidence levels across agents, and refinements that emerge over time. Moreover, they do not reliably support cross-agent comparison of preference relations. Consequently, single-level models exhibit inherent constraints when representing diverse agent viewpoints across different issue pairs. To overcome these limitations, this paper introduces a multi-level preference framework that generalizes single-level preference, converse, and indifference relations by incorporating multiple levels of relational intensity, thereby enabling a more fine-grained characterization of agents’ preference strengths over issue pairs. Within this framework, we define conflict measures for individual issue pairs between two agents, and further extend them to a set of issues, facilitating the exact quantification of conflict degrees between two agents across multiple issues, and enabling a more accurate trisection of agent pairs into alliance, neutrality, and conflict relations. As a concrete instantiation, we develop a two-level preference-based model distinguishing strong and weak relations, and apply it to a case study on development planning in Gansu Province. A comparative analysis demonstrates that the multi-level preference framework not only captures conflicts with greater expressiveness and accuracy than single-level approaches but also yields richer and more actionable insights for conflict resolution, thereby enhancing both the interpretability and the practical value of three-way conflict analysis.</div></div>","PeriodicalId":13842,"journal":{"name":"International Journal of Approximate Reasoning","volume":"190 ","pages":"Article 109621"},"PeriodicalIF":3.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145880984","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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":"Neural approaches to SAT solving: Design choices and interpretability","authors":"David Mojžíšek ,&nbsp;Jan Hůla ,&nbsp;Ziwei Li ,&nbsp;Ziyu Zhou ,&nbsp;Mikoláš Janota","doi":"10.1016/j.ijar.2025.109609","DOIUrl":"10.1016/j.ijar.2025.109609","url":null,"abstract":"<div><div>In this contribution, we provide a comprehensive evaluation of graph neural networks applied to Boolean satisfiability problems, accompanied by an intuitive explanation of the mechanisms enabling the model to generalize to different instances. We introduce several training improvements, particularly a novel closest assignment supervision method that dynamically adapts to the model’s current state, significantly enhancing performance on problems with larger solution spaces. Our experiments demonstrate the suitability of variable-clause graph representations with recurrent neural network updates, which achieve good accuracy on SAT assignment prediction while reducing computational demands. We extend the base graph neural network into a diffusion model that facilitates incremental sampling and can be effectively combined with classical techniques like unit propagation. Through analysis of embedding space patterns and optimization trajectories, we show how these networks implicitly perform a process very similar to continuous relaxations of MaxSAT, offering an interpretable view of their reasoning process. This understanding guides our design choices and explains the ability of recurrent architectures to scale effectively at inference time beyond their training distribution, which we demonstrate with test-time scaling experiments.</div></div>","PeriodicalId":13842,"journal":{"name":"International Journal of Approximate Reasoning","volume":"190 ","pages":"Article 109609"},"PeriodicalIF":3.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145735354","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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":"Weighted feature graph-based multilabel feature selection via multi-metrics with global–local correlation","authors":"Lin Sun ,&nbsp;Changwu Feng ,&nbsp;Xiankun Zhang ,&nbsp;Jiucheng Xu","doi":"10.1016/j.ijar.2025.109623","DOIUrl":"10.1016/j.ijar.2025.109623","url":null,"abstract":"<div><div>Due to the increasing prevalence of multilabel data, multilabel classification must tackle the challenges of high-dimensional feature spaces, complex label dependencies, and sample sparsity, which restricts the effectiveness of multilabel learning. To address these challenges, this work constructs a weighted feature graph-based multilabel feature selection methodology via multi-metrics with global–local correlation. First, fuzzy similarity relations in feature space and fuzzy decision similarity within label space are calculated. A joint fuzzy similarity relation is then constructed to capture the consistency of samples across both spaces. Next, we derive associativity from the joint fuzzy similarity relation, obtain redundancy from the fuzzy similarity relation, and measure interactivity using the fuzzy dependency degree of feature subsets. These three metrics are combined to define the edge weights of the feature graph, thereby describing the complex correlations among features for multilabel classification. Second, to evaluate the global–local correlation, a ridge regression coefficient matrix is presented. Concurrently, the correlation weight of information energy ratio is calculated via mutual information between features and labels, forming a feature-label correlation matrix. We then use a multi-Criteria Decision-Making (MCDM) framework to balance these global and local correlations. This allows us to develop a weighted feature-label matrix and a relative closeness degree for each alternative, which determines the node weights of the weighted feature graph. Third, to dynamically adjust the intensity of information propagation between features, a feature-driven, attention-based weight allocation strategy is studied. We construct a feature node matrix that represents multi-source information fusion and an attention matrix by combining graph structural features with the MCDM results. These are used to form a feature-aware hybrid adjacency matrix. An improved PageRank scheme then iteratively updates the feature ranking scores on this hybrid structure, to select a discriminative and representative feature subset. Experiments illustrate that our methodology outperforms other comparative approaches across several metrics on high-dimensional multilabel data.</div></div>","PeriodicalId":13842,"journal":{"name":"International Journal of Approximate Reasoning","volume":"190 ","pages":"Article 109623"},"PeriodicalIF":3.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145837549","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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":"Data speak but sometimes lie: A game-theoretic approach to data bias and algorithmic fairness","authors":"Chiara Manganini,&nbsp;Esther Anna Corsi,&nbsp;Giuseppe Primiero","doi":"10.1016/j.ijar.2025.109608","DOIUrl":"10.1016/j.ijar.2025.109608","url":null,"abstract":"<div><div>In the present work, we develop a novel information-theoretic and logic-based approach to data bias in Machine Learning predictions and show its relevance in the specific context of fairness evaluation. We frame predictions made on biased data as Ulam games, which formalise key aspects of data-driven inference, and from which a variation of the rational non-monotonic consequence relation can be defined. We investigate this framework to model how differential levels of noise in input features impact Machine Learning predictions. To the best of our knowledge, this is the first game-theoretic formalisation of ML unfairness.</div></div>","PeriodicalId":13842,"journal":{"name":"International Journal of Approximate Reasoning","volume":"190 ","pages":"Article 109608"},"PeriodicalIF":3.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145788224","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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":"How artificial intelligence leads to knowledge why: An inquiry inspired by Aristotle’s Posterior Analytics","authors":"Guus Eelink ,&nbsp;Kilian Rückschloß ,&nbsp;Felix Weitkämper","doi":"10.1016/j.ijar.2025.109603","DOIUrl":"10.1016/j.ijar.2025.109603","url":null,"abstract":"<div><div>Bayesian networks and causal models provide frameworks for reasoning about external interventions, enabling tasks that go beyond what probability distributions alone can support. Although these formalisms are often informally described as encoding causal knowledge, there is a lack of a formal theory that characterizes the kind of knowledge required to predict the effects of such interventions. This work introduces the theoretical framework of <em>causal systems</em> to implement Aristotle’s distinction between knowledge-<em>that</em> and knowledge-<em>why</em> within the setting of artificial intelligence. By interpreting existing AI technologies as causal systems, it examines the corresponding forms of knowledge they embody. Finally, it argues that predicting the effects of external interventions is possible only with knowledge-<em>why</em>, offering a more precise account of the assumptions underlying this capacity.</div></div>","PeriodicalId":13842,"journal":{"name":"International Journal of Approximate Reasoning","volume":"190 ","pages":"Article 109603"},"PeriodicalIF":3.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145788263","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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":"Connecting classical finite exchangeability to quantum theory and indistinguishability","authors":"Alessio Benavoli ,&nbsp;Alessandro Facchini ,&nbsp;Marco Zaffalon","doi":"10.1016/j.ijar.2025.109604","DOIUrl":"10.1016/j.ijar.2025.109604","url":null,"abstract":"<div><div>Exchangeability is a fundamental concept in probability theory and statistics. It allows to model situations where the order of observations does not matter. The classical de Finetti’s theorem provides a representation of infinitely exchangeable sequences of random variables as mixtures of independent and identically distributed variables. The quantum de Finetti theorem extends this result to symmetric quantum states on tensor product Hilbert spaces. It is well known that both theorems do not hold for finitely exchangeable sequences. The aim of this work is to investigate two lesser-known representation theorems, which were developed in classical probability theory to extend de Finetti’s theorem to finitely exchangeable sequences by using quasi-probabilities and quasi-expectations. With the aid of these theorems, we illustrate how a de Finetti-like representation theorem for finitely exchangeable sequences can be formulated through a mathematical representation which is formally equivalent to quantum theory (with boson-symmetric density matrices). We then show a promising application of this connection to the challenge of defining entanglement for indistinguishable bosons.</div></div>","PeriodicalId":13842,"journal":{"name":"International Journal of Approximate Reasoning","volume":"189 ","pages":"Article 109604"},"PeriodicalIF":3.0,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145681530","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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":"Multi-granularity Knowledge Fusion for Feature Selection Using Granular-ball Entropy Uncertainty Measures","authors":"Kehua Yuan ,&nbsp;Yuji Bai ,&nbsp;Duoqian Miao ,&nbsp;Weiping Ding ,&nbsp;Yiyu Yao ,&nbsp;Hongyun Zhang ,&nbsp;Witold Pedrycz","doi":"10.1016/j.ijar.2025.109590","DOIUrl":"10.1016/j.ijar.2025.109590","url":null,"abstract":"<div><div>Multi-granularity computing for knowledge discovery has emerged as a remarkable paradigm in data mining and machine learning. As a representative method, granular-ball computing has attracted considerable attention due to its efficiency and adaptability in handling complex data distributions. However, most existing granularity-based approaches focus on intra-granular mutual information while neglecting the heterogeneity and overlapping phenomena across granularities. This limitation often leads to imprecise knowledge space construction and inaccurate uncertainty estimation in feature evaluation. To overcome this problem, this study proposes a novel and high-efficiency multi-granularity knowledge fusion framework for feature selection, incorporating an enhanced granular-ball generation mechanism and a newly designed granular-ball entropy (GB-E) uncertainty measure. Specifically, we first develop an enhanced granular-ball generation mechanism to construct multi-granularity knowledge space by incorporating class distribution information, thus achieving more accurate and flexible data partitioning. Subsequently, by jointly analyzing the separation and aggregation among granular balls, a novel granular-ball entropy is proposed to quantify uncertainty in the multi-granularity knowledge space. Compared with existing uncertainty measure methods, it provides a dual-perspective uncertainty characterization and effectively improves the accuracy of granularity information fusion. Furthermore, two feature significance measures based on the proposed GB-E measure are introduced for feature evaluation, and then a corresponding feature selection method is developed. Extensive experiments on multiple public datasets demonstrate the proposed method’s superior classification performance compared with several state-of-the-art approaches.</div></div>","PeriodicalId":13842,"journal":{"name":"International Journal of Approximate Reasoning","volume":"189 ","pages":"Article 109590"},"PeriodicalIF":3.0,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145621172","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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":"Machine learning for quantifier selection in cvc5","authors":"Jan Jakubův ,&nbsp;Mikoláš Janota ,&nbsp;Jelle Piepenbrock ,&nbsp;Josef Urban","doi":"10.1016/j.ijar.2025.109602","DOIUrl":"10.1016/j.ijar.2025.109602","url":null,"abstract":"<div><div>In this work we considerably improve the real-time performance of state-of-the-art SMT solving on first-order quantified problems by efficient machine learning guidance of quantifier selection. Quantifiers represent a significant challenge for SMT and are technically a source of undecidability. In our approach, we train an efficient machine learning model that informs the solver which quantifiers should be instantiated and which not. Each quantifier may be instantiated multiple times and the set of the currently active quantifiers changes as the solving progresses. Therefore, we invoke the ML predictor many times, during the whole run of the solver. To make this efficient, we use fast ML models based on gradient boosted decision trees. We integrate our approach into the state-of-the-art cvc5 SMT solver and show a considerable increase of the system’s holdout-set performance after training it on large sets of first-order problems. The method is tested in several ways, using both single-strategy and portfolio approaches. The evaluation is done on two large formal verification corpora: first-order problems created from the Mizar Mathematical Library, and first-order problems created from the HOL4 standard library.</div></div>","PeriodicalId":13842,"journal":{"name":"International Journal of Approximate Reasoning","volume":"189 ","pages":"Article 109602"},"PeriodicalIF":3.0,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145621175","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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 robust multi-source transfer classification method based on belief functions for cross-domain pattern recognition","authors":"Linqing Huang ,&nbsp;Jinfu Fan ,&nbsp;Gongshen Liu ,&nbsp;Shilin Wang","doi":"10.1016/j.ijar.2025.109589","DOIUrl":"10.1016/j.ijar.2025.109589","url":null,"abstract":"<div><div>In pattern recognition with few or no labeled data, domain adaptation techniques are often used to transfer knowledge in the source domain to help build classification models in the target domain. The effective combination of complementary information in multiple source domains usually can further improve the classification accuracy. To this end, we present a Robust Multi-Source Transfer (RMST) classification method consisting of two-step fusion of classifiers to extract the useful information in each source domain as much as possible, and to effectively combine the complementary information using belief functions in different source domains. The first step is to accurately predict the pseudo labels when computing conditional distribution, in order to learn a robust new feature representation, which is used to fuse diverse classifiers learnt by patterns for reliable soft classification in the second step of our algorithm. Furthermore, the soft classification results yielded by the assistance of different source domains are combined by belief functions with the new weighting factors taking into account the distribution discrepancy and classifier’s performance. The effectiveness of RMST was evaluated with respect to a variety of advanced methods, and the experimental results show that RMST can significantly improve the classification accuracy.</div></div>","PeriodicalId":13842,"journal":{"name":"International Journal of Approximate Reasoning","volume":"189 ","pages":"Article 109589"},"PeriodicalIF":3.0,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145464523","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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":"ARIPOTER: Solvers for approximate reasoning based on grounded semantics","authors":"Jérôme Delobelle ,&nbsp;Jean-Guy Mailly ,&nbsp;Julien Rossit","doi":"10.1016/j.ijar.2025.109599","DOIUrl":"10.1016/j.ijar.2025.109599","url":null,"abstract":"<div><div>Efficient computation of hard reasoning tasks is a key issue in abstract argumentation. One recent approach is to define approximate algorithms, <em>i.e.</em> methods that provide an answer that may not always be correct, but outperform the exact algorithms regarding the computation runtime. One such approach proposes to use the grounded semantics, which is polynomially computable, as a starting point for determining whether arguments are (credulously or skeptically) accepted with respect to various other extension-based semantics. In this paper, we push further this idea by defining a general family of approaches to evaluate the acceptability of arguments which are not in the grounded extension, neither attacked by it. These approaches rely on gradual semantics to evaluate these arguments. We also propose an approach using an heuristic based on the number of arguments attacked by or attacking an argument, and we show that this last approach, although seemingly different, is actually also an instance of our general family of approaches based on gradual semantics. We have implemented our approaches and provided an empirical study in which we discuss the results and compare our approach with the state-of-the-art approximate algorithms.</div></div>","PeriodicalId":13842,"journal":{"name":"International Journal of Approximate Reasoning","volume":"189 ","pages":"Article 109599"},"PeriodicalIF":3.0,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145577154","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}