Mean-field theory of learning dynamics in deep neural networks

Time：2023-10-10 Tue 10:00-12:00

Venue：A3-1a-205 ZOOM: 537 192 5549（PW: BIMSA） Abstract

Organizer：Seyed Hamidreza Mofidi, Shailesh Lal, Hossein Yavartanoo

Speaker：Cengiz Pehlevan SEAS, Harvard University

Abstract

Learning dynamics of deep neural networks is complex. While previous approaches made advances in mathematical analysis of the dynamics of two-layer neural networks, addressing deeper networks have been challenging. In this talk, I will present a mean field theory of the learning dynamics of deep networks and discuss its practical implications.

DATEOctober 10, 2023

Related News

0
Learning constitutive models with neural networks
AbstractIn this talk, I will introduce some work of learning constitutive equations in fluid mechanics and geophysics based on machine learningSpeaker Intro熊繁升，现任北京雁栖湖应用数学研究院助理研究员，曾任北京应用物理与计算数学研究所所聘博士后。先后毕业于中国地质大学(北京)、清华大学，美国耶鲁大学联合培养博士。研究兴趣主要集中于基于机器学习算法(DNN、PINN、DeepONet等)求解微分方程模型正/反问题...
1
A Neural Networks Solver for R Matrices
Abstractwe describe how neural networks may be used to learn solutions to the Yang-Baxter Equation using 2d spin chains of difference form as a concrete example. the talk is based on arxiv:2304.07247, which is joint work with Suvajit Majumder and Evgeny Skvortsov.Speaker IntroDr Shailesh Lal received his PhD from the Harish-Chandra Research Institute. His research interests are applications of ...