KAIST 정보보호대학원에서는 12월 3일 오후 4시에 아래와 같이 콜로퀴움을 개최하고자 합니다. 많은 참석 부탁드립니다. 

o 일 시 : 24. 12. 3(화) 16:00~
o 주 제 : What Does Automatic Differentiation Compute for Non-Differentiable Functions?
o 강 사 : 이원열 교수님 (포항공대 교수)
o 장 소 : N1동 201호
​
※ 시작시간 5분전에 준비하여 주세요.

ㅡㅡㅡ

♣ 제목 : What Does Automatic Differentiation Compute for Non-Differentiable Functions?
 
♣ 강연정보
Automatic differentiation (AD) is a class of algorithms for computing the derivatives of given functions. It lies at the core of many areas in computer science, including machine learning and scientific computing. While AD was originally developed for differentiable functions over continuous domains (e.g., real numbers), in practice it is often applied to non-differentiable functions (e.g., neural networks using ReLU) over discrete domains (e.g., floating-point numbers). This raises a natural question: is AD correct in any formal sense when applied to such non-differentiable functions over discrete domains? In this talk, I will present some answers to this question, based on joint work with my colleagues (ICLR 2024 Spotlight, ICML 2023, NeurIPS 2020 Spotlight). The talk will likely be delivered in Korean.
 
♣ 약력
Wonyeol Lee is an Assistant Professor of Computer Science at POSTECH. Before joining POSTECH, he was a Postdoctoral Associate at CMU, working with Feras Saad. He received his PhD degree in Computer Science from Stanford University, advised by Alex Aiken. During his PhD, he also spent time at KAIST for military service, working with Hongseok Yang. He obtained his BS degree in Computer Science and Mathematics from POSTECH.