Speaker: 金睿楠博士，中国科学院香港创新研究院
Inviter: 张波、程涤非
Title: A Comprehensive Framework for Analyzing the Convergence of Adam: Bridging the Gap with SGD
Language: Chinese
Time & Venue: 2025.1.7 10:00-11:00 N208（#腾讯会议：766-569-933）
Abstract: Adaptive Moment Estimation (Adam) is an important optimization algorithm in deep learning. However, despite its practical success, the theoretical understanding of Adam's convergence has been constrained by stringent assumptions, such as almost surely bounded stochastic gradients or uniformly bounded gradients, which are more restrictive than those typically required for analyzing Stochastic Gradient Descent (SGD). In this talk, we introduce a novel and comprehensive framework for analyzing the convergence properties of Adam.
附件下载：
