Artificial Intelligence Review最新文献

{"title":"Feeding, grooming, dressing, and body repositioning: categorizing four pillars of learning-based manipulation for robotic caregiving","authors":"Wenkai Chen,&nbsp;Shifeng Huang,&nbsp;Hui Zhang,&nbsp;Yuyang Tu,&nbsp;Yao Lu,&nbsp;Qingdu Li,&nbsp;Jianwei Zhang","doi":"10.1007/s10462-026-11524-7","DOIUrl":"10.1007/s10462-026-11524-7","url":null,"abstract":"<div>\u0000 \u0000 <p>Robotic caregiving is emerging as a critical application domain where human-centered robotics, learning-based manipulation, and embodied intelligence converge. As global aging accelerates, assistive robots are expected to support essential daily activities, such as feeding, grooming, dressing, and body repositioning, which demand precise, adaptive, and safe physical interaction with users. Despite rapid progress, these tasks remain challenging due to deformable objects, diverse human behaviors, and safety-critical, contact-rich dynamics. This survey provides the first unified review of learning-based robotic caregiving across all four core manipulation tasks. We analyze recent advances through a cross-cutting framework grounded in four major learning paradigms: multimodal perception, imitation learning, reinforcement learning, and embodied reasoning. For each caregiving task, we synthesize how these paradigms address unique perceptual, physical, and safety requirements, identifying shared principles and task-specific characteristics. We also examine simulation platforms as enablers for scalable training and reproducible evaluation. This survey further summarizes a conceptual unified perception-reasoning-control loop that distills common structures across caregiving tasks. Finally, we highlight open challenges, including user-centered personalization, safety validation, and ethical deployment, and outline research directions toward trustworthy and adaptive assistive robots.</p>\u0000 </div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 4","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11524-7.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440966","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 privacy-preserving federated learning for intrusion detection systems","authors":"Thomas Bunko,&nbsp;Michael N. Johnstone,&nbsp;Wencheng Yang,&nbsp;Ben A. Scott","doi":"10.1007/s10462-026-11519-4","DOIUrl":"10.1007/s10462-026-11519-4","url":null,"abstract":"<div><p>Intrusion detection systems (IDS) monitor and detect malicious activity and unauthorized access that may compromise systems. Traditional IDS approaches send data to a central server for analysis, raising privacy concerns as data owners lose control over security. Federated Learning (FL) offers a privacy-preserving alternative by allowing local devices to process their data and generate models without sharing raw data. These local models are aggregated centrally to form a comprehensive model with performance comparable to centralized systems. This paper reviews FL-based IDS research, and is the first review paper to focus on privacy-preserving techniques collectively known as privacy-preserving Federated Learning (PPFL) for IDS. We examine methods used to prevent data leakage while maintaining detection effectiveness, including encryption-based and lightweight alternatives. While FL keeps raw data local, it remains susceptible to inference and poisoning attacks. Our findings show that most FL-based IDS research concentrates on data locality alone, with limited adoption of additional privacy-enhancing techniques. Advancing PPFL-IDS requires moving beyond data while addressing trade-offs. This review highlights key gaps and directions for future research.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 5","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11519-4.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147562075","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 on design choices for self-supervised learning in computer vision","authors":"Ladyna Wittscher","doi":"10.1007/s10462-026-11506-9","DOIUrl":"10.1007/s10462-026-11506-9","url":null,"abstract":"<div><p>Self-supervised learning is a successful strategy to overcome data scarcity and improve robustness in computer vision by adding a pretext task that can exploit inherent data relationships as supervision signals during pretraining. However, the combination of pretraining and downstream training renders model design more complex, as additional design choices are required. This paper analyses the effects of such design choices specific to self-supervised learning on model performance and robustness. How does the pretext task influence the downstream task and how to design an ideal and generalizable pretext task? Which properties of the pretraining dataset are favorable and how similar should the pretext and downstream dataset ideally be? To address these questions, a comprehensive survey has been conducted, encompassing the results of diverse models and publications with different design choices. The results demonstrate the advantages of in-domain pretraining and the importance of aligning all design choices in order to ensure optimal results. Furthermore, the characteristic differences between predictive, contrastive and generative self-supervised learning and the design choices which are crucial for each of these learning paradigms are analyzed in detail.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 4","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-02-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11506-9.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147441932","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 unified framework for training deep neural networks with better generalization via self-ensemble label correction","authors":"Yuelong Xia,&nbsp;Wenjie Liu,&nbsp;Xiaodi Sun,&nbsp;Jing Yang,&nbsp;Yihang Tong,&nbsp;Yungang Zhang","doi":"10.1007/s10462-026-11523-8","DOIUrl":"10.1007/s10462-026-11523-8","url":null,"abstract":"<div><p>The generalization capabilities of deep neural networks have been significantly improved by applying a wide spectrum of label modification approaches, e.g., label smoothing, confidence penalty, label correction, etc. However, less attention has been paid to label correction. In this paper, we propose self-ensemble label correction (SEELC), a unified training framework that dynamically calibrates and distills own knowledge to leverage the training process for better generalization. We analyze the generalization of SEELC from three different perspectives: 1) Our analysis shows that SEELC not only alleviates model miscalibration but also improves model robustness to random noise and adversarial noise, highlighting that SEELC enhances the generalization under supervised learning settings; 2) Our analysis shows that on the power of pseudo-label and noisy student, SEELC can be easily extended to semi-supervised learning and effectively handle domain discrepancies in unsupervised domain adaptation (UDA); 3) Our analysis also sheds light on understanding self-supervised learning, e.g., avoiding degenerate solutions, and can be well explained from alignment and uniformity. Finally, experiments on three applications show the superiority of our approach, i.e., classification on clean and noisy data, UDA and linear evaluation protocol in self-supervised learning.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 4","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-02-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11523-8.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147561033","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":"Recent trends of machine learning techniques for risk assessment in hazardous environments","authors":"Nesma El-Sokkary,&nbsp;A. A. Arafa,&nbsp;E. G. Zahran,&nbsp;Hesham A. Hefny,&nbsp;Nagy Ramdan","doi":"10.1007/s10462-026-11507-8","DOIUrl":"10.1007/s10462-026-11507-8","url":null,"abstract":"<div><p>Risk assessment is a critical step in the regulatory decision-making process, carried out within the framework of political and legislative requirements, in addition to the need to make decisions on time according to the available resources. Some critical and hazardous facilities such as nuclear power plants, offshore oil and gas, and hazardous materials storage sites, are very useful to society but are inherently risky. For these facilities, failure has an increased criticality, causing adverse effects on the ecological system and human health. Therefore, the risk assessment process is time-sensitive for such industries. Due to the recent technological development in the industry, the significance of risk management has increased, and the identification, assessment, reporting, and management of risks have received continuous attention. Machine learning is becoming more and more powerful for use in industry applications; many solutions have already been put into practice, and many more are being investigated. Most articles do not review the hazard industries. This review aims at identifying and analyzing the literature on risk assessments for the study of risks, types of consequences, and disaster mitigation, with a focus on literature that uses machine learning approaches, particularly in hazard environments. Retrieved articles are analyzed and reviewed in terms of different risk assessment aspects. Findings and gaps in each article are reported. The results of the analysis prove the power of machine learning approaches in assessing the risk and highlight their use in hazardous environments. Findings also showed that it is an ongoing research topic that needs more studies to achieve the highest benefits. Besides, this review can provide researchers with the future directions in this field.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 3","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11507-8.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147340553","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":"From classical machine learning to emerging foundation models: review on multimodal data integration for cancer research","authors":"Amgad Muneer,&nbsp;Muhammad Waqas,&nbsp;Maliazurina B. Saad,&nbsp;Eman Showkatian,&nbsp;Rukhmini Bandyopadhyay,&nbsp;Hui Xu,&nbsp;Wentao Li,&nbsp;Joe Y. Chang,&nbsp;Zhongxing Liao,&nbsp;Cara Haymaker,&nbsp;Luisa Solis Soto,&nbsp;Carol C. Wu,&nbsp;Natalie I. Vokes,&nbsp;Xiuning Le,&nbsp;Lauren A. Byers,&nbsp;Don L. Gibbons,&nbsp;John V. Heymach,&nbsp;Jianjun Zhang,&nbsp;Jia Wu","doi":"10.1007/s10462-026-11522-9","DOIUrl":"10.1007/s10462-026-11522-9","url":null,"abstract":"<div><p>Cancer research is increasingly driven by the integration of diverse data modalities, spanning from genomics and proteomics to imaging and clinical factors. However, extracting actionable insights from these vast and heterogeneous datasets remains a key challenge. The rise of foundation models (FMs) large deep-learning models pretrained on extensive amounts of data serving as a backbone for a wide range of downstream tasks—offers new avenues for discovering biomarkers, improving diagnosis, and personalizing treatment. This paper presents a comprehensive review of widely adopted integration strategies of multimodal data to assist advance the computational approaches for data-driven discoveries in oncology. We examine emerging trends in machine learning (ML) and deep learning (DL), including methodological frameworks, validation protocols, and open-source resources targeting cancer subtype classification, biomarker discovery, treatment guidance, and outcome prediction. This study also comprehensively covers the shift from traditional ML to FMs for multimodal integration. We present a holistic view of recent FMs advancements and challenges faced during the integration of multi-omics with advanced imaging data. We identify state-of-the-art FMs, publicly available multi-modal repositories, and advanced tools and methods for data integration. We argue that current state-of-the-art integration methods provide the essential groundwork for developing the next generation of large-scale, pre-trained models poised to further revolutionize oncology. To the best of our knowledge, this is the first review to systematically map the transition from conventional ML to advanced FM for multimodal data integration in oncology, while also framing these developments as foundational for the forthcoming era of large-scale AI models in cancer research. The GitHub repo of this project available at https://github.com/WuLabMDA/Medical-Foundation-Models.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 4","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11522-9.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147441482","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":"Artificial intelligence in Traditional Chinese Medicine herbs: a survey","authors":"Jinrun Wu,&nbsp;Qiang He,&nbsp;Dong Yang,&nbsp;Yuliang Cai,&nbsp;Jiawei Zhang,&nbsp;Xi Lu,&nbsp;Hua Zhu","doi":"10.1007/s10462-026-11513-w","DOIUrl":"10.1007/s10462-026-11513-w","url":null,"abstract":"<div><p>Traditional Chinese Medicine (TCM) herbs are central to both prevention and treatment, yet their complex mechanisms remain only partly understood. This review examines how artificial intelligence (AI) has been applied to advance TCM herbal research, tracing its evolution from expert systems to machine learning, deep learning, and large language models. By analyzing both the traditional approach, rooted in classical TCM principles, and the modern approach, informed by biomedical science, we identify how AI enables knowledge extraction, prescription recommendation, compound discovery, and mechanism exploration. The review highlights the unique contribution of AI in bridging empirical heritage with scientific rigor, while also outlining key challenges such as data standardization and ethical considerations. Our findings demonstrate that the integration of AI with TCM herbs not only preserves traditional knowledge but also accelerates innovation in personalized and integrative medicine.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 4","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11513-w.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147441481","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":"QRL-AFOFA: Q-learning enhanced self-adaptive fractional order firefly algorithm for large-scale and dynamic multiobjective optimization problems","authors":"Yashar Mousavi,&nbsp;Parastoo Akbari,&nbsp;Rashin Mousavi,&nbsp;Arash Mousavi,&nbsp;Ibrahim Beklan Kucukdemiral,&nbsp;Afef Fekih,&nbsp;Umit Cali","doi":"10.1007/s10462-026-11511-y","DOIUrl":"10.1007/s10462-026-11511-y","url":null,"abstract":"<div>\u0000 \u0000 <p>This paper introduces QRL-AFOFA, a Q-learning-enhanced adaptive fractional-order firefly algorithm developed to address the challenges of large-scale and dynamic multiobjective optimization problems. While fractional-order metaheuristics provide memory-driven search dynamics and reinforcement learning (RL) offers adaptive policy control, existing hybrid methods often face critical limitations such as parameter sensitivity, premature convergence, and poor diversity preservation. To overcome these challenges, QRL-AFOFA integrates five synergistic innovations: real-time adaptive tuning of fractional-order parameters, entropy-regularized Q-value updates, stagnation-aware restart strategies, reflection-based boundary handling, and dual-phase learning rate scheduling. The Q-learning framework autonomously adapts critical parameters while entropy regularization maintains the exploration-exploitation balance, and stagnation-aware mechanisms ensure the preservation of population diversity. Extensive experiments on the IEEE Congress on Evolutionary Computation (CEC2021) benchmark functions demonstrate that QRL-AFOFA consistently outperforms state-of-the-art algorithms across diverse problem categories. Statistical validation further confirmed its superior performance across multiobjective, large-scale, and dynamic optimization scenarios. The algorithm achieves exceptional performance in high-dimensional settings while eliminating manual parameter tuning requirements, positioning it as an intelligent, scalable optimization framework for complex real-world applications.</p>\u0000 </div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 4","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11511-y.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147441474","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":"Lies, damned lies, and language statistics: a comprehensive review of risks from manipulation, persuasion, and deception with large language models","authors":"Cameron Jones,&nbsp;Benjamin Bergen","doi":"10.1007/s10462-026-11517-6","DOIUrl":"10.1007/s10462-026-11517-6","url":null,"abstract":"<div><p>Large Language Models (LLMs) have the potential to produce content that is effective at persuading, deceiving, and manipulating people. Here we survey the possible risks of systems with these capabilities, including criminal fraud, political misinformation, addictive AI companions, and misaligned autonomous systems. We then survey the rapidly growing body of empirical work on their propensity to deceive and their capacity to persuade, which suggests that models are already roughly as persuasive as untrained human participants. We review proposed mitigations for these techniques—including training models to be truthful or monitoring their hidden states—and highlight strengths and weaknesses of each potential approach. Finally, we highlight five key open questions for future research: how persuasive could AI systems be? How do AI systems persuade? What broader social impacts could AI persuasion have? Does persuasion advance truth? And how effective are proposed mitigations?</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 4","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11517-6.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147441475","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":"Enhancing the prediction of heating and cooling loads in residential buildings using explainable XGBoost: a comparative analysis of SHAP and LIME techniques","authors":"Meysam Alizamir,&nbsp;Sungwon Kim,&nbsp;Salim Heddam,&nbsp;Aliakbar Gholampour,&nbsp;Hyounseok Moon","doi":"10.1007/s10462-026-11521-w","DOIUrl":"10.1007/s10462-026-11521-w","url":null,"abstract":"<div><p>Accurate prediction of building thermal demands is essential for evaluating energy efficiency and optimizing performance within energy management systems. This need has become increasingly critical due to growing environmental concerns surrounding energy consumption and its ecological impact. Reliable estimation of energy demands supports informed decision-making in resource management and carbon footprint reduction, promoting sustainable and technologically advanced construction practices. This study investigates the predictive accuracy of several ensemble boosting algorithms in estimating thermal energy demands (heating and cooling loads) using different architectural parameters. The performance evaluation using RMSE metrics demonstrated XGBoost’s dominance across both thermal prediction tasks. For heating load estimation, this algorithm yielded the lowest error rate of 0.309 kW, surpassing CatBoost’s 0.318 kW and LightGBM’s 0.351 kW results. In cooling load forecasting, XGBoost maintained its leading position with 0.731 kW RMSE, outperforming LightGBM (0.907 kW) and CatBoost (0.917 kW), which occupied the subsequent ranking positions. Testing phase findings across heating and cooling load predictions confirmed that the XGBoost ensemble methodology successfully improved estimation accuracy when applied to thermal demand modeling with associated parameters. Moreover, SHAP analysis revealed that glazing area distribution and orientation had minimal impact on heating load predictions, while overall height and relative compactness were the most influential. Similarly, orientation and glazing area distribution parameters contribute the least to the prediction of cooling load, whereas overall height and relative compactness parameters have the most significant influence. These findings confirm the effectiveness of XGBoost in enhancing prediction accuracy for residential thermal demand modeling.</p><h3>Graphical abstract</h3><div><figure><div><div><picture><source><img></source></picture></div></div></figure></div></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"59 4","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-026-11521-w.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147441476","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}