Jyh-Shing Roger Jang, National Taiwan Univ. (NTU), CTO E.Sun Financial Holding Company.
We have been building faster computers to host innovative ML (machine learning) applications such as image recognition and language understanding. But can we reverse the roles and use ML to help EDA (electronic design automation) to create better performing chips
and computers? The answer is definitely yes. In this talk, we shall cover two such examples, including wafer failure analysis and side channel
attacks using ML. We shall address the guidelines for ML corpus
design, and touch on corpus-hungry ML applications that outperform humans by a large margin and were unthinkable before the current
AI era.