[2023.06.12(Mon.)]Artificial Intelligence & AI Convergence Network Colloquium

[행사] [2023.06.12(Mon.)]Artificial Intelligence & AI Convergence Network Colloquium

When : 2023년 6월 12일(월), 오후 1시 30분

Where : 팔달관 407호

Speaker : 유재준 교수(UNIST AI 대학원 & 전기전자공학)

Title : Can We Find Strong Lottery Tickets in Generative Models?

Abstract : Yes. In this talk, I show that it is possible to find a strong generative model

without training, a.k.a the existence of strong lottery tickets, the subnetworks that achieve good

performance without any weight update. Neural network pruning is considered the main

cornerstone of model compression for reducing the costs of computation and memory.

Unfortunately, pruning a generative model has not been extensively explored, and all existing

pruning algorithms suffer from excessive weight-training costs, performance degradation,

limited generalizability, or complicated training. To address these problems, we propose to find a

strong lottery ticket via moment-matching scores. Our experimental results show that the

discovered subnetwork can perform similarly or better than the trained dense model even when

only 10% of the weights remain. To the best of our knowledge, we are the first to show the

existence of strong lottery tickets in generative models and provide an algorithm to find it

stably.

Bio : 2021.07 - 현재 UNIST AI 대학원 전임 & 전기전자공학 겸임 조교수

2019.12 - 2021.07 EPFL Postdoctoral researcher

2018.02 - 2019.12 NAVER Search&Clova AI research scientist

2018.02 KAIST 박사 바이오및뇌공학과 / 지도교수 예종철

Host : 소프트웨어학과 손경아 교수(kasohn@ajou.ac.kr)

AI Convergence Network