时间：2024年10月28日（星期一）下午14:55-17:30

Time：14:55-17:30, Monday, October 28th, 2024

主持人：理学院物理系PI鲍芳琳

Host: Prof. Fanglin Bao, PI of School of Science, Westlake University

地点：西湖大学云谷校区E10-222

Venue: E10-222, Yungu Campus, Westlake University

讲座语言：中文（Prof. Pan Zhang)／英文 (Prof. Henri Orland)

Lecture Language: Chinese (Prof. Pan Zhang) / English (Prof. Henri Orland)

   Prof. Pan Zhang, Professor,        

Institute of Theoretical Physics, Chinese Academy of Sciences   

Prof. Henri Orland, Senior theoretical physicist,

Institut de Physique Théorique, Université Paris-Saclay, CEA, Saclay

主讲人/Speaker:

Pan Zhang is a professor at the Institute of Theoretical Physics, Chinese Academy of Sciences, and the director of the Second Research Division. He conducts research at the intersection of statistical physics, quantum physics, and machine learning. Together with collaborators, Pan Zhang proposed the sparse-state tensor network method to solve the sampling problem of Google's Sycamore quantum circuits. With collaborators, he also introduced the VAN framework for statistical mechanics and the MPS Born machine, a tensor network machine learning model. His research have been supported by the 2023 National Science Fund for Distinguished Young Scholars and have earned him the 2023 Chinese Academy of Sciences Young Scientist Award, as well as the second prize of the 2022 Beijing Natural Science Award. Since 2023, Pan Zhang has served as an editorial board member of Physical Review Letters (DAE in the field of physics and machine learning).

Henri Orland is a senior theoretical physicist at the Institut de Physique Théorique, CEA, Saclay. His research interests encompass quantum many-body theory & nuclear physics, statistical physics of disordered systems, biological systems, and soft condensed matter. Orland's work has been recognized with the CNRS Bronze Medal in 1981 and the Légion d'Honneur in 2011. In 2023, he was elected as a member of the American Academy of Arts and Sciences. Orland has written more than 200 research papers published in major journals; he holds the patent for an algorithm simulating protein folding and is the author of two books, Quantum Many-Particle Systems with John W. Negele and Molecular Kinetics in Condensed Phases: Theory, Simulation, and Analysis with Ron Elber and Dmitri Makarov.

摘要/Abstract:

The 2024 Nobel Prize in Physics was awarded to John Hopfield and Geoffrey Hinton in recognition of their foundational discoveries and innovations in neural network-based machine learning. In this talk, Pan Zhang will explore these contributions from the perspective of statistical physics, beginning with the Ising model and gradually introducing the key contributions of Hopfield and Hinton. These include the Hopfield model, the backpropagation algorithm, Boltzmann machines, unsupervised pre-training, and the AlexNet deep neural network. Additionally, the talk will review the remarkable collaboration between statistical physics and machine learning at the end of the last century and provide a brief outlook on the future directions of physics and machine learning.

We show how the Optimal Transport problem can be recast in the framework of statistical physics and studied at finite temperature. This allows for the implementation of very fast and efficient algorithms and the generalization of the Wasserstein distance to finite temperature. The statistical physics method can be extended to several forms of unbalanced Optimal Transport problems, and to the Gromov-Wasserstein metric. We show applications to shape recognition as well as to some biological problems. Finally, we discuss some implications in non-equilibrium statistical physics.

讲座联系人/Contact:

School of Science, Zeyuan LI, Email: lizeyuan@westlake.edu.cn

School of Science, Kaiyin XU, Email: xukaiyin@westlake.edu.cn