Interdisciplinary Institute in Artificial Intelligence

DESCRIPTION

I am interested in interdisciplinary research at the intersection of statistical physics and machine learning. Today, machine learning is becoming a powerful tool for physicists to analyze various physics problems. Conversely, the concepts and techniques developed in physics help us uncover the mystery behind the outstanding performance of machine learning architectures, which have traditionally been viewed as black boxes. We are currently developing machine learning methods inspired by statistical physics, fostering synergy between physics and machine learning.

Generative models are having a huge impact on society, such as image generation and ChatGPT, to name just a few. Although they have achieved remarkable success, the interpretability of their outstanding performance remains poorly understood. Instead, they are generally regarded as black boxes. We are conducting interdisciplinary research at the intersection of statistical physics and generative models to understand the mechanisms of learning and to develop efficient generation algorithms. In particular, he is developing generative models based on a renormalization group technique. The renormalization group is a powerful technique in modern theoretical physics that allows us to process data hierarchically on a scale-by-scale basis. This new generative model enables us to interpret both the learning process and the results in a physically intuitive way, and provides us with a new, fast Monte Carlo sampling algorithm to generate new examples.

CHAIR ACTIVITIES

Entropy plays an important role in science and engineering, as it characterizes various features of systems, such as the time evolution of a macroscopic sample, phase transitions, the emergence of certain orders, and so on. However, measuring entropy is a challenging task, particularly when the system is out of equilibrium. This stems mainly from the difficulty of obtaining a high-dimensional probability distribution from data (e.g., experimental images, simulated configurations). We are developing a method to measure Shannon entropy using the generative model mentioned above. The method will be applied to image data from various fields, including physics, chemistry, and biology.