The convergence of Generative Artificial Intelligence (GenAI) and Affective Computing (AffComp) opens new frontiers in creation of emotionally intelligent systems, with growing relevance for human–machine interaction across domains such as virtual assistants, avatars, gaming, and personalised digital experiences. In this paper, we explore the opportunities arising from the integration of GenAI with AffComp across various application areas—data augmentation and synthesis, emotion generation, personalised content creation, data imputation, and emotion transfer and style manipulation—with a particular focus on the deployment of emotionally intelligent systems through a risk–reward balancing framework. The study adopts an empirical and evaluative approach, leveraging state-of-the-art GenAI models within a multimodal AffComp framework to demonstrate significant improvements in emotional expressiveness, adaptive behavior, and personalized responses, enhancing human–computer interaction in applications such as virtual assistants, avatars, and gaming. However, this convergence introduces privacy, ethical, and security risks. Mitigating these risks requires privacy-preserving techniques, real-time bias detection, transparent and culturally diverse models, strong ethical frameworks, user education, and interdisciplinary collaboration to support responsible deployment, ongoing evaluation, and robust oversight. By balancing the associated risks and rewards, we outline strategies to maximise the benefits of GenAI and AffComp while mitigating the potential harms. This paper invites the research community to collaborate on shaping the future of emotionally intelligent AI systems, prioritising both innovation and the well-being of individuals and society as a whole.
Large language models (LLMs) have demonstrated impressive in-context learning capabilities, particularly in few-shot prompting. However, most studies have focused on natural language processing and computer vision, applications to complex physiological signals such as electroencephalogram (EEG) decoding remain largely unexplored. This paper proposes purity-guided active prompting (PGAP), which integrates domain-specific EEG feature extraction with an active learning based demonstration selection strategy. PGAP identifies a compact set of prototypical examples that effectively activate the few-shot capabilities of LLMs without gradient-based parameter updates, and is compatible with both open-source and closed-source LLMs. Unlike the conventional framework that involves random demonstration selection on a per-instance basis for each test sample, PGAP performs a one-time identification of highly representative EEG samples from the sample pool, significantly reducing computational overhead. Extensive experiments on seven EEG datasets and three paradigms demonstrated that PGAP consistently outperformed five existing sample selection baselines, achieving superior accuracy and robustness. Furthermore, the performance improved with more powerful LLMs (e.g., DeepSeek-V3, GPT-4.1), and sometimes even surpassed state-of-the-art supervised models.
The convergence of demographic shifts, increasing demand for continuous care, and sustained pressure on healthcare systems is accelerating the transition toward the Internet of Medical Things (IoMT). At the core of this transformation is wireless sensing—a contactless paradigm that leverages ambient radio-frequency (RF) signals to infer physiological and behavioral information. By exploiting these signals, it becomes possible to continuously monitor vital signs and human activities without the compliance burden associated with wearable devices.
Neural Architecture Search (NAS) offers a promising pathway to automate the design of deep neural networks, yet it faces a significant challenge in balancing computational efficiency with search effectiveness in a huge search space. Coarse-to-fine (C2F) NAS methods mitigate this by narrowing the architecture search to promising subspaces. However, by focusing only on a single region, they risk overlooking globally optimal architectures due to the multi-modal nature of the search space. This paper highlights the limitation of existing C2F approaches and motivates the need for more sophisticated search strategies capable of efficiently exploring multiple promising regions to achieve a better performance-efficiency trade-off. To achieve this goal, we propose a multi-stage NAS (MstageNAS) framework that progressively constructs multiple high-quality subspaces and implements an efficient exploration within them. MstageNAS initiates with a coarse search to identify promising architectures within the whole search space. Subsequently, the individual subspace is constructed around each promising architecture. To ensure the quality of this subspace, an architecture explanation method is devised to identify the core sub-structure of the promising architecture and use it to form the basis of the individual subspace. Finally, a Monte Carlo-based search strategy is developed to facilitate architecture search within these subspaces, with the goal of striking a good balance between exploration and exploitation. We evaluate the proposed MstageNAS framework across search spaces of various types and tasks. Extensive experiments demonstrate that MstageNAS can outperform state-of-the-art NAS methods or achieve comparable performance but with about 2× less search cost.
Due to factors such as changes in customer demand and supply disruption, design changes during product production are often inevitable. In many cases, a change to a shared key part (also known as component) has an impact on multiple related products owned by a company, requiring simultaneous design changes for these products. To simultaneously generate the optimal change plans for these affected products (i.e., series products), this paper proposes a multi-task multi-objective evolutionary generation algorithm based on part similarity. Firstly, a multi-task multi-objective optimization model is constructed on the basis of the multi-layer complex network model designed for series products. In this model, multiple indicators including the product service performance impact, carbon emissions, change costs and duration are considered. Subsequently, a multi-task multi-objective evolutionary algorithm based on part similarity is developed to solve the above model. Several new operators/strategies including the part similarity evaluation measure, the multi-task knowledge transfer strategy, and the multi-task constraint handling mechanisms are designed to enhance the algorithm’s performance. Finally, the effectiveness of both the proposed model and the proposed algorithm is validated on the design change problem of multiple TVs in Skyworth RGB Co., Ltd. Experimental results show that the proposed algorithm can generate highly competitive optimal change plans for multiple TVs simultaneously.
Video-based analysis of infant movement has emerged as a promising approach for the early detection of neurodevelopmental disorders. The field has evolved from subjective clinical assessment to integrated, quantitative, and automated systems. This comprehensive survey charts the technological evolution of the field from 2014 to 2024, focusing on three core areas: pose estimation methods adapted for infant subjects, direct video analysis approaches, and dataset collection methodologies. The survey systematically analyzes the progression from basic motion tracking to advanced deep learning solutions, with particular attention to key developmental periods, including writhing movements, fidgety movements, and voluntary movements. The review highlights both significant advances and persistent challenges. Successful adaptations of pose estimation techniques to infant characteristics contrast with unresolved issues in data acquisition, privacy preservation, and clinical integration. A growing emphasis on explainable artificial intelligence is essential for fostering clinical trust and adoption. A comparative analysis of datasets and monitoring systems clarifies their respective strengths and limitations. Key challenges include the scarcity of large and diverse datasets, the lack of standardization across platforms, and the need for robust clinical validation studies. The findings suggest that while technical capabilities have advanced considerably, successful clinical implementation requires careful consideration of practical constraints and ethical issues. This survey serves as a comprehensive reference for researchers and clinicians working at the intersection of computer vision, machine learning, and pediatric healthcare.
Juergen Branke is currently a Professor of Operational Research and Systems at Warwick Business School, University of Warwick, U.K. His research specializes in the interface of optimization, machine learning, and simulation, with a specific focus on decision-making under uncertainty, dynamically changing environments, and multi-objective optimization.
Davide Bacciu received the Ph.D. degree in computer science from IMT Lucca, Italy, for which he was awarded the 2009 E.R. Caianiello prize. Currently, he is with the University of Pisa, Italy, where he is a Full Professor and the founder and head of the Pervasive AI Laboratory. His research interests include neural networks, generative models, learning for graphs, continual learning, and distributed and embedded learning systems; in these fields, he has authored more than 200 peer-reviewed articles.
Evolutionary multitasking optimization (EMTO) has emerged as a powerful tool for addressing complex optimization problems, particularly those involving multiple related tasks. However, current EMTO methods prioritize promoting convergence characteristics while paying little attention to potential privacy leakage issues during knowledge transfer among tasks. This paper designs privacy-preserving centralized and decentralized EMTO models based on inter-task knowledge transfer, and proposes a perturbation query strategy for the centralized EMTO. Furthermore, passive and active attacks in EMTO are proposed for the first time, and an example of active attacks in EMTO is provided by constructing a deceptive malicious task. The active attack technique can serve as a test to determine whether the EMTO methods can effectively identify and defend against malicious knowledge transfer. Experimental results show that the proposed privacy-preserving centralized EMTO method can protect clients’ privacy information while maintaining algorithm performance, and the proposed active attack technique can significantly slow down the convergence speed of the target task.
Class imbalance tends to result in biased classifiers that favor the majority class and disadvantage the minority class. Unfortunately, the minority class is usually of crucial importance in many real-life applications. Hybrid sampling methods address this issue by oversampling the minority class to increase the number of its instances, followed by undersampling to remove low-quality instances. However, most existing sampling methods face difficulties in generating diverse high-quality instances and often fail to remove noise or low-quality instances on a larger scale effectively. Therefore, this paper proposes an evolutionary multi-granularity hybrid sampling method, called EvoSampling, to tackle this hindrance. During the oversampling process, genetic programming is used with multi-task learning to effectively and efficiently generate diverse high-quality instances. During the undersampling process, a granular ball-based undersampling method is employed to remove noise in a multi-granular fashion, thereby enhancing data quality. Experiments on 18 imbalanced datasets demonstrate that EvoSampling effectively enhances the performance of various classification algorithms by providing better datasets than existing sampling methods. Ablation studies further indicate that allowing knowledge transfer accelerates the evolutionary learning process.
As a kind of high order neural network, Sigma-Pi-Sigma has strong nonlinear mapping ability, but the complexity of its network structure often leads to low application efficiency and lack of theoretical results. Although traditional regularization methods are more effective in sparsity and preventing overfitting, they do not take into account the topology and connection information between the data. Therefore, in order to sparsely optimize the network structure for improved efficiency and to enable rigorous theoretical analysis of complex inter-node relationships, graph-structured data must be processed more effectively. In this paper, the graph regularization component is integrated into the Sigma-Pi-Sigma neural network (SPSNN) learning framework, and the SPSNN online gradient algorithm based on the Turlaplacian regularization is studied, considering the entropy error measure. Under reasonable conditions, the boundedness of the weights, the monotonicity of the error function, and the convergence of the algorithm are strictly proved. Subsequently, the effectiveness of this method is verified through extensive experimental data. The simulation results are in good agreement with the theoretical results.
Klaus-Robert Müller is Full Professor for Machine Learning at the Department of Computer Science at Technische Universität Berlin and at the Department of Artificial Intelligence at Korea University, Seoul. Over the years he held the roles as director of the Bernstein Center for Neurotechnology, co-director of the Berlin Center for Big Data and director of the Berlin Machine Learning Center. In 2021, he became director of the Berlin Institute for Foundations of Learning and Data (BIFOLD)—one of 5 permanently funded German national centers for AI. In 2020/2021 and recently in 2024/2025 he was on a short (1-year-long) sabbatical from academia to lead a team at Google Brain rsp. DeepMind as Principal Researcher. In 2012, he was elected to be a member of the German National Academy of Sciences—Leopoldina; in 2017 of the Berlin Brandenburg Academy of Sciences; in 2022 member of the German National Academy of Engineering; and also, in 2017 an external scientific member of the Max-Planck Society (MPII). He received several research awards: among others, in 2014 the Berlin Science Prize awarded by the governing mayor of Berlin; in 2017 the Vodafone Innovation Award; in 2024 the Hector Science award and the Feynman Award; and in 2025 the IEEE CIS Neural Network Pioneer Award. Consecutively from 2019 on he became an ISI Highly Cited Researcher. His research interests are in the field of machine learning, deep learning, explainable AI, and data analysis covering a wide range of theory and numerous scientific (physics, chemistry and medicine) and industrial applications. Google Scholar Citations > 168000, h-index 167.
The IEEE Computational Intelligence Society (CIS) offers many financially-sponsored conferences each year. All of these depend on the hard work and effort of volunteers who propose and then organize these events in the pursuit of a healthy exchange of information and improvement in the theory and application of computational intelligence approaches. The IEEE CIS annually publishes a list of the general chairs, program chairs, and financial chairs of IEEE CIS financially-sponsored conferences in the IEEE Computational Intelligence Magazine as a way of recognizing these volunteers for their service. Your efforts are definitely appreciated and thank you on behalf of the society!
