Abstract
This article argues that the integration of machine learning (ML) techniques with well-established algorithmic approaches can overcome challenges faced by current applications of ML to electronic design automation problems. The standard cell routing problem is used as a showcase for the proposed approach.