Nov 16, 2023 Thursday
Abstract:
Online learning is an important learning paradigm for dealing with sequential prediction and decision-making problems. The non-stationarity issue is a central challenge for modern online learning, given that many real-world data streams are collected in open environments and the data distributions are naturally changing over time. This talk will present our recent efforts on this topic. We introduce dynamic regret as the performance measure to guide the algorithm design and propose the “online ensemble” framework to optimize the measure. We will demonstrate that the framework can yield fruitful algorithmic and theoretical results for several important online learning problems and attain best-known or optimal guarantees.
Time:
Nov 16, 2023 Thursday
16:30-17:20
Location:
RmE1-101, GZ Campus
Zoom:
628 334 1826 (PW: 234567)
Bilibili Live:
ID: 30748067
Speaker Bio:
Dr. Peng Zhao
Assistant Professor, AI school, Nanjing University
Peng Zhao is an assistant professor in the AI school at Nanjing University. His research explores the theoretical foundations of machine learning, with a focus on online learning, stochastic optimization, and open-environment machine learning. His work has been published at top-tier conferences and journals such as ICML/NeurIPS/COLT and JMLR. He serves regularly as the reviewer/area chair for various top-tier conferences and journals, and he served as the workflow chair for AAAI 2019.