Program Synthesis meets AI
Guest Speaker: Armando Solar-Lezama
In this talk, I describe some of the recent work by my group at the intersection of Program Synthesis with AI and machine learning. First, I will talk about some recent work on leveraging ideas and techniques from Machine Learning to make Program Synthesis more effective, including the question of how we can learn better abstractions to improve program synthesis given a corpus of synthesis problems. The talk will also describe how program synthesis can be a useful tool for problems that require generalization from small numbers of examples and interpretability of the learned functions.
Armando Solar-Lezama is an associate professor at MIT CSAIL, where he leads the Computer-Aided Programming group. The group focuses on developing techniques and tools that exploit automated reasoning and large amounts of computing power to tackle challenging programming problems.