招待講演1 [11月12日 (水) 13:30 – 14:30]
Invited Talk 1 [Wednesday, November 12, 1:30 – 2:30 p.m.]
Nicolò Cesa-Bianchi (Università degli Studi di Milano / Politecnico di Milano, Italy)
Multiagent Online Learning
We study a network of agents that learn by interacting with the environment and with the other agents. In the talk, we show the extent to which communication allows us to prove performance bounds that are strictly better than the known bounds for non-communicating agents. Our results are formulated within the online learning setting and hold either when agents learn a single common task or when each agent learns a different task.
招待講演2 [11月13日 (木) 12:30 – 13:30]
Invited Talk 2 [Thursday, November 13, 0:30 – 1:30 p.m.]
Koaru Ota (Tohoku University / Muroran Institute of Technology)
AI-Powered Next-Generation Communication and Applications
The evolution toward 6G networks requires intelligent and resilient communication systems. This talk introduces two AI-driven approaches: RIS-assisted 6G networks, which optimize resource allocation based on traffic prediction to enhance user experience, and UAV-based semantic communication for disaster response, enabling efficient and robust emergency connectivity. These studies demonstrate how AI can address bandwidth limitations, adapt to diverse requirements, and support critical applications for future communication beyond 5G.
招待講演3 [11月14日 (金) 13:15 – 14:15]
Invited Talk 3 [Friday, November 14, 1:15 – 2:15 p.m.]
Yu-Chiang Frank Wang (National Taiwan University / NVIDIA)
Improved Efficiency & Reasoning for Vision-Language Models
The convergence of vision, language, and generative modeling is driving the next wave of AI applications. In this talk, we explore how recent advances in efficiency and reasoning are enabling more practical and powerful vision-language models (VLMs). We begin by surveying new distillation techniques that significantly reduce the size and inference cost of VLMs without compromising their multimodal reasoning capabilities. We then examine emerging strategies for multimodal reasoning. We will discuss how these developments are reshaping the landscape of visual reasoning, opening the door to scalable, adaptive, and user-centric multimodal systems.
招待講演4 [11月14日 (金) 14:15 – 15:15]
Invited Talk 4 [Friday, November 14, 2:15 – 3:15 p.m.]
Leander Thiele (Kavli IPMU)
Machine Learning for the next chapter in Cosmology
Cosmology uses observations of the largest scales in the Universe to understand the origin, composition, and evolution of the cosmos. We are currently experiencing an explosion in size and quality of our data sets. This massively increased information should enable us to advance our knowledge of the Universe to unprecedented levels. The data reduction and analysis challenge, however, is formidable. Machine learning has emerged as a promising tool that will be instrumental in meeting this challenge. After introducing the state of the field, I will describe some applications of machine learning in cosmology and outline a roadmap for the path forward.