Cat-Ho NGUYEN (Duy Tan University, Vietnam)

Title: TBA

Abstract.

Bio

Shogo Okada (Japan Advanced Institute of Science and Technology, Japan)

Title: Social Multimodal Intelligence: From Human Understanding to Social Understanding and Decision Making

Abstract. Human behavior and social interaction contains rich information about cognitive states, emotions, intentions, interpersonal relationships, and collective social dynamics. Recent advances in multimodal sensing and machine learning have enabled the extraction of such information from speech, language, facial expressions, body movements, and physiological signals. These developments are opening new opportunities for understanding not only individuals but also social interactions and decision-making processes.
In this talk, I introduce the emerging concept of Social Multimodal Intelligence (SMI), a research framework that integrates multimodal behavioral sensing, machine learning, and computational modeling to bridge human understanding, social understanding, and intelligent decision support. First, I present methods for estimating individual cognitive and psychological characteristics, including personality traits, emotions, engagement, and early signs of cognitive decline from multimodal behavioral signals. Next, I discuss approaches for modeling social phenomena in dyadic and group interactions, such as rapport formation, communication skills, and group performance. These studies demonstrate how latent social structures can be inferred from observable multimodal behaviors. Building upon these foundations, I introduce intelligent systems that utilize social understanding for decision making and adaptation, including adaptive conversational agents, healthcare support systems, and interview training system. Such systems continuously interpret multimodal social signals and dynamically optimize their behavior according to users’ attitute and states.
Finally, I discuss future challenges and opportunities in Social Multimodal Intelligence, including multimodal foundation models, social reasoning, human-AI collaboration, and socially intelligent decision support.

Bio

Shogo Okada is a Professor at the Japan Advanced Institute of Science and Technology (JAIST), where he leads the Social Signal Processing and Multimodal Interaction Group. He has been a Professor since 2024, following his appointment as an Associate Professor at JAIST in 2017. He received his Ph.D. from the Tokyo Institute of Technology in 2008 and has held academic positions at the Tokyo Institute of Technology, Kyoto University, and the Idiap Research Institute, Switzerland, where he served as a Visiting Faculty Member in 2014. His research focuses on Social Multimodal Intelligence, Social Signal Processing, multimodal interaction, human behavior understanding, and machine learning. His work aims to bridge human understanding, social understanding, and intelligent decision-making through the analysis of speech, language, behavior, and physiological signals. Dr. Okada has received several prestigious awards, including the 2025 Best Paper Award from the Japanese Society for Artificial Intelligence (JSAI), the JSAI 30th Anniversary Best Paper Award, the Best Paper Award at HCI International 2019, three Best Paper Runner-up Awards at ACM ICMI 2019 and ICMI2022. He is a member of ACM, IEEE, JSAI, and IEICE.

Cassio de Campos (Eindhoven University of Technology, Netherlands)

Title: The Relationship between Tensor Factorizations and Circuits

Abstract. This talk presents two important machine learning approaches: circuit representations and tensor factorizations. They belong to seemingly distinct yet fundamentally related areas. The work generalizes popular tensor factorizations within the circuit language, and unifies various circuit learning algorithms under a single, generalized hierarchical factorization framework. We discuss about a modular “Lego block” approach to build tensorized circuit architectures, allowing us to systematically construct and explore various circuit and tensor factorization models while maintaining tractability. This connection not only clarifies similarities and differences in existing models, but also enables the development of pipelines for building and optimizing these machine learning models in a more interpretative and reliable manner.

Bio

Cassio de Campos a full professor and head of the Uncertainty in AI group at TU Eindhoven. He obtained his degrees from the University of Sao Paulo (Brazil) in Computer Science and Mechatronics. He is a Senior Member of ACM and board member of the Association for Uncertainty in AI. He works on foundations of artificial intelligence and machine learning, in particular related to robust and interpretative machine learning models, including probabilistic generative models, imprecise probability, and computational complexity. Cassio has published more than 150 scientific papers, and supervised numerous grad and undergrad students. He is active in research projects with Dutch NWO and EU funding. He served as reviewer and panelist of multiple research foundations and (senior) area chair of many major AI and machine learning (ML) conferences, as well as area editor of IJAR and senior associate editor of ACM Transactions on Probabilistic ML. He organised and chaired multiple events including UAI and PGM, and is appointed general (co-)chair of ECMLPKDD 2027.

Naoki Masuyama (Osaka Metropolitan University, JAPAN)

Title: Adaptive Resonance Theory-based Clustering and its Extensions

Abstract. Clustering is a fundamental technique for extracting meaningful structure from data and has been applied in a wide variety of fields. Self-organizing growing clustering methods, such as Growing Neural Gas (GNG) and Self-Organizing Incremental Neural Networks (SOINN), can adaptively capture the geometric structure of data through topological networks composed of nodes and edges. However, these methods are not always well suited to continual learning scenarios, where new information must be incorporated without destroying previously acquired knowledge. Adaptive Resonance Theory (ART)-based clustering provides a principled framework for addressing this challenge through a balance between stability and plasticity. This talk introduces the fundamentals of ART-based clustering and discusses its recent extensions.

Bio

Naoki Masuyama received the B.E. degree in Aerospace Engineering from Nihon University, Funabashi, Japan, in 2010, the M.E. degree in Human Mechatronics Systems from Tokyo Metropolitan University, Hino, Japan, in 2012, and the Ph.D. degree in Computer Science from the Faculty of Computer Science and Information Technology, University of Malaya, Kuala Lumpur, Malaysia, in 2016. He was a Postdoctoral Research Fellow at the University of Malaya from 2016 to 2017. From October 2017 to March 2022, he was an Assistant Professor with the Graduate School of Engineering, Osaka Prefecture University, Sakai, Japan. After the university integration in 2022, he continued as an Assistant Professor with the Department of Core Informatics, Graduate School of Informatics, Osaka Metropolitan University, Sakai, Japan, from April 2022 to September 2022. Since October 2022, he has been an Associate Professor with the Department of Core Informatics, Graduate School of Informatics, Osaka Metropolitan University.
His research interests include clustering, data mining, and continual learning, with a particular focus on supervised and unsupervised continual learning methods that adaptively and efficiently extract useful information from dynamically changing environments. He has also worked on applications of continual learning to related areas such as evolutionary computation and explainable machine learning. He has authored more than 100 peer-reviewed papers, including publications in IEEE Transactions on Pattern Analysis and Machine Intelligence, IEEE Transactions on Neural Networks and Learning Systems, and IEEE Transactions on Evolutionary Computation. His work has been recognized by several awards, including the Springer Best Paper Award at EMO 2019, the Best Paper Award at IFSA 2023, and the Emerging Research Leader Award from the Japan Society for Fuzzy Theory and Intelligent Informatics (SOFT) in 2024.

Jessada Karnjana (National Electronics and Computer Technology Center, Thaidland)

Title: The Rising Threat of Voice Spoofing: Evolution of Detection in the Era of AI-Generated Speech

Abstract. Automatic Speaker Verification (ASV) has become a cornerstone of modern biometric authentication, used in applications ranging from high-security access control to everyday financial services. However, this growing adoption has been matched by fast-growing sophistication in spoofing attacks. A wide range of spoofing attacks, from impersonation and replay to AI-generated and deepfake speech, are driving an unprecedented arms race in voice security.
This talk reviews the evolution of spoofing detection, from early feature engineering to modern deep learning approaches. It covers spoofing-aware ASV system design and key evaluation frameworks such as Equal Error Rate (EER) and detection cost functions, which are essential for benchmarking performance. It also examines emerging threats, including adversarial attacks that challenge the robustness of current models.
The talk further presents insights from our recent research, including studies on spectral features such as MFCC, LFCC, and GTCC, as well as analysis-oriented descriptors like timbre and shimmer. These results highlight both the strengths and limitations of current spoofing countermeasures. By connecting fundamental concepts with recent advances, this keynote provides a clear perspective on the state of voice biometrics and outlines a forward-looking roadmap for securing digital identity in the age of generative AI.

Bio

Dr. Jessada Karnjana is the Head of the Core AI Research Team at the National Electronics and Computer Technology Center (NECTEC), National Science and Technology Development Agency (NSTDA), Thailand. He also serves as an adjunct faculty member at the Sirindhorn International Institute of Technology (SIIT), Thammasat University. He holds two Ph.D. degrees in Engineering and Technology from SIIT and in Information Science from the Japan Advanced Institute of Science and Technology (JAIST).
Dr. Karnjana has led numerous international research collaborations across the ASEAN region, serving as a principal investigator for initiatives supported by ASEAN COSTI, the e-Asia Joint Research Program, and ASEAN IVO. His earlier work focused on environmental protection and disaster prevention, particularly through the development of real-time monitoring systems using wireless sensor networks, remote sensing, and data fusion. His current research continues to evolve, spanning biometrics, AI in healthcare, and AI for environmental sustainability, where he integrates signal processing and machine learning to address complex real-world challenges.

Masaaki Ida (National Institution for Academic Degrees and Quality Enhancement of Higher Education, JAPAN)

Title: Some Topics on Fundamentals of Data Science

Abstract. In recent years, data science has been rapidly spreading to various fields of our society. This is particularly common in practical fields related to the social sciences, such as business, finance, and education. Generative AI has advanced remarkably, and it is possible to obtain sufficiently practical outputs from the system. Many modern generative AI systems are based on a large-scale neural network architecture called Transformer. This presentation will begin by introducing examples of generative AI application using LLMs (document data integrity assessment and knowledge acquisition through fine-tuning). When utilizing these methods, understanding the system— "why it works well"—is an issue that must not be ignored. This means ensuring the validity, reliability, and explainability of the machine learning model. Theoretical approaches have been taken to pursue explainability. However, it remains difficult to fully understand the entire mechanism mathematically. On the other hand, progress was made in understanding the basic functions of the system, including data transformation and learning methods. For example, the double descent phenomenon related to bias-variance and neural tangent kernels have been studied as recent important research topics. This presentation will provide an overview of these topics with related keywords including large-scale parameters, randomness, nonlinearity, kernel, and spectral analysis, and aims to contribute to the advancement of research on Integrated Uncertainty in Knowledge Modelling and Decision Making.

Bio

Prof. Masaaki Ida received the B.E., M.E, and Ph.D. degrees, all in engineering, from Kyoto University, Kyoto, Japan. He is currently a professor of National Institution for Academic Degrees and Quality Enhancement of Higher Education, Japan. He was a board member of Japan Society for Fuzzy Theory and Intelligent Informatics (SOFT) four times as editor-in-chief, academic award committee chair, archives committee chair, and vice president of SOFT. His current research interests are in computer science, especially in advanced data science and database on higher education. He has been responsible for works related to university evaluation and quality enhancement of higher education, as well as the related system development of numerous practical information systems.