Artificial Intelligence Review最新文献

{"title":"Gray langurs optimizer: a multi-group bio-inspired optimization algorithm","authors":"Saeid Barshandeh,&nbsp;Nima Khodadadi,&nbsp;Benyamin Abdollahzadeh,&nbsp;Ali Mohammadzadeh,&nbsp;El-Sayed M. El-Kenawy,&nbsp;Marwa M. Eid,&nbsp;Khalid M. Mosalam","doi":"10.1007/s10462-026-11529-2","DOIUrl":"10.1007/s10462-026-11529-2","url":null,"abstract":"<div>\u0000 \u0000 <p>Many real-world problems are optimizable in nature. These problems often have numerous decision variables and constraints, making it difficult to optimize their objective function. To solve these problems, a group of algorithms known as metaheuristic algorithms has been introduced, which, considering the problem constraints, can find the optimal values of the decision variables in a reasonable amount of time to optimize the value of the objective function(s). Due to the importance and high application of these algorithms, a new metaheuristic optimization algorithm called Gray Langurs Optimizer (GLO) has been introduced in this research. The fundamental inspiration for the GLO is the group behavior of gray langurs in nature. The gray langurs are observed in three groups: one-male, multi-male, and all-male groups. The GLO mathematically simulates the group behavior and social hierarchy of gray langurs in all three groups. The migration of gray langurs between groups due to the death of the alpha or other group members, mate finding, and puberty is formulated by GLO. Additionally, the GLO models the arbitrary behavior of gray langurs within their territory. The GLO is tested on twenty-three classical test functions, including unimodal, multimodal, and fixed-dimension benchmarks, as well as twenty-seven CEC17 test functions, which comprise shifted, rotated, and composite benchmarks. The GLO is also applied to six real-world applications, and its results are compared with those of eight state-of-the-art algorithms both numerically and visually. The experimental results demonstrate the high capability of GLO in optimizing the various test functions and real-world applications. The MATLAB and Python implementations of GLO are available at https://nimakhodadadi.com.</p>\u0000 </div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 7","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-04-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11529-2.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148237310","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deconstructing deep imbalanced regression: a comprehensive review and experimental evaluation","authors":"Noah C. Puetz,&nbsp;Jens U. Brandt,&nbsp;Marc Hilbert,&nbsp;Elena Raponi,&nbsp;Thomas Bäck,&nbsp;Thomas Bartz-Beielstein","doi":"10.1007/s10462-026-11570-1","DOIUrl":"10.1007/s10462-026-11570-1","url":null,"abstract":"<div><p>In real-world applications, there is a fundamental problem: the data most critical to predict interesting events, anomalies, and high-stakes outliers are the rarest, while less interesting data is abundant. Although deep learning is deployed specifically for these difficult prediction tasks, data-driven models inevitably fail in underrepresented areas. This discrepancy between the empirical data- and the desired evaluation distribution is equivalent to a target distribution shift. The research field, termed Deep Imbalanced Regression (DIR), has emerged explicitly to address this challenge, which is particularly acute for continuous targets where most conventional classification-based methods are ill-suited. In this paper, we present the first comprehensive review of the DIR landscape, organized around a novel two-axis taxonomy that disentangles challenges along a <i>Data Axis</i> (target distribution shift, continuity, and density) and a <i>Deep-Learning Axis</i> (shared capacity, biased updates, and manifold distortion), where the latter captures a cascading failure mechanism through which deep models systematically neglect underrepresented targets. Within this framework, we systematically categorize and analyze 19 state-of-the-art methods spanning architectural, algorithm-level, and representation learning approaches, and empirically re-evaluate twelve of them with publicly available implementations under controlled, identical conditions. To stress-test generalization across the full target range, we introduce three novel targeted evaluation protocols, <i>Balanced Extrapolation</i>, <i>Bimodal Interpolation</i>, and <i>Blind-Spot Isolation</i>, that expose failure modes hidden by standard benchmarks (https://github.com/noah-puetz/deconstructing_deep_imbalanced_regression). Our study underscores the significant impact of imbalance on regression accuracy, offering a conceptual framework and practical benchmarks to catalyze further development of systems capable of capturing the rare as reliably as the common.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 6","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-04-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11570-1.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147754380","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Trust as a design principle in human–robot collaboration: a review of explainable and adaptive control","authors":"Abdelrahman T. Elgohr,&nbsp;Maher Rashad,&nbsp;Eman M. El-Gendy,&nbsp;Waleed Shaaban,&nbsp;Mahmoud M. Saafan","doi":"10.1007/s10462-026-11554-1","DOIUrl":"10.1007/s10462-026-11554-1","url":null,"abstract":"<div>\u0000 \u0000 <p>Human–Robot Collaboration (HRC) has emerged as a fundamental element of new industrial and service systems, wherein humans and robots function within common physical and cognitive environments to achieve shared goals. In addition to traditional issues of safety and productivity, trust has become a critical element influencing cooperation efficiency, human dependence, and sustained system acceptability. This review offers a thorough and reliability-focused summary of HRC research, highlighting the significance of explainable intelligence and adaptive control in promoting trustworthy collaboration. The study initially identifies trust as a fundamental design target in HRC, delineating its dynamic, multifaceted characteristics and its impact on human decision-making and interaction behavior. A systematic review methodology is employed to analyze cutting-edge approaches across essential dimensions, including trust modeling and estimation, multimodal human state and intention recognition, explainable artificial intelligence (XAI) techniques, and adaptive and learning-based control architectures. The analysis emphasizes the role of transparency, interpretability, and context-aware adaptation in establishing trust calibration within safety-critical and dynamic collaborative environments. A cross-sectional synthesis of the literature reveals several critical gaps, including the lack of standardized trust evaluation metrics, limited integration of explainability and control adaptation, inadequate consideration of long-term trust dynamics, and insufficient validation in real-world, unstructured environments. The analysis closes by delineating potential research avenues for cohesive, human-centered HRC frameworks that effortlessly incorporate trust modeling, explainable decision-making, and adaptive control. The ideas offered seek to inform the creation of advanced collaborative robots that are safe, efficient, transparent, adaptive, and trustworthy.</p>\u0000 </div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 5","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-04-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11554-1.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147738254","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A critical review of state-of-the-art explainable artificial intelligence (XAI) methods and their business applications","authors":"Tingshen Lin,&nbsp;Raymond Y. K. Lau,&nbsp;Shenghui Hu","doi":"10.1007/s10462-026-11565-y","DOIUrl":"10.1007/s10462-026-11565-y","url":null,"abstract":"<div><p>With the growing adoption of AI in high-risk business sectors such as finance and healthcare, ensuring transparency, explainability, and controllability of AI decision-making has become increasingly critical. Traditional “black box” models, particularly deep learning and large language models, pose challenges to trust, compliance, and operational reliability. In response, explainable artificial intelligence has emerged as a key research area bridging technological innovation and practical business needs. This article systematically examines the classification, mechanisms, and evolution of mainstream XAI methods, with a focus on both technological development and sustainable business applications. Guided by two core research questions, we analyze model-specific and model-agnostic advancements, summarize approaches in terms of faithfulness, cost-effectiveness, stability, and operational applicability, and analyze their adaptability in key industries such as finance and healthcare. Finally, we identify current challenges and future development trends in XAI, offering practical recommendations for building trustworthy, transparent, and sustainable business AI systems.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 7","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-04-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11565-y.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148173085","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"IMPACTX: improving model performance by appropriately constraining the training with teacher explanations","authors":"Andrea Apicella,&nbsp;Salvatore Giugliano,&nbsp;Francesco Isgró,&nbsp;Andrea Pollastro,&nbsp;Roberto Prevete","doi":"10.1007/s10462-026-11564-z","DOIUrl":"10.1007/s10462-026-11564-z","url":null,"abstract":"<div>\u0000 \u0000 <p>The eXplainable Artificial Intelligence (XAI) research predominantly concentrates on providing explanations about AI model decisions, especially Deep Learning (DL) models. However, there is a growing interest in using XAI techniques to automatically improve the performance of the AI systems themselves. This paper proposes <i>IMPACTX</i>, a novel approach that leverages XAI as a fully automated attention mechanism, without requiring external knowledge or human feedback. Experimental results show that IMPACTX improves predictive performance compared to standalone baseline models by integrating XAI-based supervision provided by teacher explanations into the training process. Furthermore, IMPACTX directly provides new feature attribution maps for the model’s decisions, without relying on external XAI methods during the inference process. Our proposal is evaluated using three widely recognized DL models (EfficientNet-B2, MobileNet, and LeNet-5) and on six publicly available benchmark datasets, comprising three image datasets (CIFAR-10, CIFAR-100, and STL-10) and three tabular datasets (Covertype, CDC Diabetes Health Indicators, and Adult). The results show that IMPACTX consistently improves the performance of all the inspected DL models across all evaluated datasets, and it directly provides appropriate explanations for its responses.</p>\u0000 </div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 7","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-04-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11564-z.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148011972","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Neural architecture search from a natural language processing perspective: a survey","authors":"Yu Yan,&nbsp;Liping Mo,&nbsp;Yangbin Tan","doi":"10.1007/s10462-026-11550-5","DOIUrl":"10.1007/s10462-026-11550-5","url":null,"abstract":"<div>\u0000 \u0000 <p>Neural architecture search has revolutionized automated machine learning and demonstrated transformative potential across natural language processing domains. This survey presents a comprehensive task-oriented analysis of NAS applications in NLP, systematically categorizing research into three core areas of text representation and classification, sequence modeling and generation, and information extraction. We trace the field’s evolution through three distinct developmental phases from foundational methodology transfer and task-specific architectural customization to deployment-oriented optimization. The analysis examines key methodological advances in search space construction, search strategy optimization, and efficient evaluation mechanisms, while highlighting architectural innovations including pre-trained model compression, multimodal fusion, and specialized designs for machine translation, speech recognition, and entity extraction. Building on this comprehensive synthesis of current progress, we identify critical future research directions in large language model optimization, zero-cost evaluation methods, and knowledge-aware architecture design for practical NLP systems.</p>\u0000 </div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 7","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11550-5.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148011875","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A survey of zero-knowledge proof based verifiable machine learning","authors":"Zhizhi Peng,&nbsp;Chonghe Zhao,&nbsp;Taotao Wang,&nbsp;Guofu Liao,&nbsp;Zibin Lin,&nbsp;Yifeng Liu,&nbsp;Bin Cao,&nbsp;Long Shi,&nbsp;Qing Yang,&nbsp;Shengli Zhang","doi":"10.1007/s10462-026-11557-y","DOIUrl":"10.1007/s10462-026-11557-y","url":null,"abstract":"<div><p>As machine learning technologies advance rapidly across various domains, concerns over data privacy and model security have grown significantly. These challenges are particularly pronounced when models are trained and deployed on cloud platforms or third-party servers due to the computational resource limitations of users’ end devices. In response, zero-knowledge proof (ZKP) technology has emerged as a promising solution, enabling effective validation of model performance and authenticity in both training and inference processes without disclosing sensitive data. Thus, ZKP ensures the verifiability and security of machine learning models, making it a valuable tool for privacy-preserving AI. Although some research has explored the verifiable machine learning solutions that exploit ZKP, a comprehensive survey and summary of these efforts remains absent. This survey paper aims to bridge this gap by reviewing and analyzing all the existing Zero-Knowledge Machine Learning (ZKML) research from June 2017 to August 2025. We begin by introducing the concept of ZKML and outlining its ZKP algorithmic setups under three key categories: verifiable training, verifiable inference, and verifiable testing. Next, we provide a comprehensive categorization of existing ZKML research within these categories and analyze the works in detail. Furthermore, we explore the implementation challenges faced in this field and discuss the improvement works to address these obstacles. Additionally, we highlight several commercial applications of ZKML technology. Finally, we propose promising directions for future advancements in this domain.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 7","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-04-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11557-y.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148011874","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A survey of graphical models for decision-making: integrating causality and game theory","authors":"Maarten C. Vonk,&nbsp;Mauricio Gonzalez-Soto,&nbsp;Anna V. Kononova","doi":"10.1007/s10462-026-11537-2","DOIUrl":"10.1007/s10462-026-11537-2","url":null,"abstract":"<div><p>Causality and game theory are two influential fields that contribute significantly to decision-making in various domains. Causality defines and models causal relationships in complex policy problems, while game theory provides insights into strategic interactions among stakeholders with competing interests. Integrating these frameworks has led to significant theoretical advancements with the potential to improve decision-making processes. However, practical applications of these developments remain underexplored. To support efforts toward implementation, this paper surveys key concepts in game theory and causality that are essential to their intersection, particularly within the context of probabilistic graphical models. By rigorously examining these concepts and illustrating them with intuitive, consistent examples, we clarify the required inputs for implementing these models, provide practitioners with insights into their application and selection across different scenarios, and reference existing research that supports their implementation. We hope this work encourages broader adoption of these models in real-world scenarios.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 7","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-04-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11537-2.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148011877","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Generative Artificial Intelligence Models: A Survey","authors":"Manal Almuammar,&nbsp;Khulud Alsultan,&nbsp;Yara Altukhaim","doi":"10.1007/s10462-026-11546-1","DOIUrl":"10.1007/s10462-026-11546-1","url":null,"abstract":"<div><p>Generative Artificial Intelligence (GenAI) has seen a remarkable growth in recent years, expanding beyond traditional discriminative tasks to the generation of new content, such as text, images, and videos, based on patterns discovered in the existing data. Advances in deep learning, particularly the development of novel architectures such as Transformers, Generative Adversarial Networks, Diffusion Models, and Variational Autoencoders have significantly contributed to this progress. GenAI has rapidly evolved, demonstrating advanced capabilities across a wide range of real-world applications such as healthcare, and creative industries. In this paper, we explore the current foundations of GenAI to provide a comprehensive overview of its principal models and their variants, highlighting their use cases, strengths, and limitations. This survey introduces an architecture-based structured taxonomy of GenAI models, laying the groundwork for the development of more efficient generative applications and encouraging further advancements within the field.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 7","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-04-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11546-1.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148173124","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Explainable AI for glaucoma detection and classification: a comprehensive review","authors":"Ahmed M. Abd El-Gawad,&nbsp;Sarah Hassan,&nbsp;Mohamed Elsharkawy,&nbsp;Shahad Al Hamadani,&nbsp;Aliyah Shivel,&nbsp;Tracy Couch,&nbsp;Ibrahim Saleh,&nbsp;Eman A. Atallah,&nbsp;Mohammed Ghazal,&nbsp;Guruprasad Giridharan,&nbsp;Hanan M. Amer,&nbsp;Abeer Twakol Khalil,&nbsp;Ayman El-Baz","doi":"10.1007/s10462-026-11551-4","DOIUrl":"10.1007/s10462-026-11551-4","url":null,"abstract":"<div><p>Glaucoma is a leading cause of irreversible blindness and often advances without symptoms. This PRISMA/PICOS–guided systematic review synthesizes 100 studies (to January 2025) on artificial intelligence (AI) for glaucoma detection, staging, and monitoring using color fundus photography (CFP), optical coherence tomography (OCT), and OCT angiography (OCTA). Roughly 45% of studies use CFP, 55–60% use OCT, and 10–15% use OCTA; about one quarter fuse multiple imaging types with clinical data. Across methods—convolutional neural networks, Transformers, and hybrid fusion—internal test results are strong, with several systems reporting specialist-level sensitivity and specificity. Translation to practice remains limited by small, single-center cohorts, device/domain shifts that reduce generalization, a lack of prospective trials, and explainability approaches that are misaligned with clinical reasoning. We recommend multi-center, multi-ethnic, multi-vendor benchmarks, rigorous external and prospective validation, clinically meaningful and standardized explainability, and cost- and workflow-aware deployment. Addressing these gaps could enable earlier detection, more personalized monitoring, and real-world impact in reducing preventable blindness. Within this review, we investigate key challenges revolving around the use of AI models for glaucoma imaging across different learning paradigms, improving model efficiency, and coupling them with complementary techniques and multimodal data (CFP, OCT, and OCTA). We hope this review can give a comprehensive picture of AI-based methods for glaucoma to readers in the field of medical image analysis.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 7","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-04-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11551-4.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148011876","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}