PRESENTATION ABSTRACT:

Neurosymbolic programming (NSP) is an emerging area of computing that bridges the fields of deep learning and program synthesis. Like in classical machine learning, the goal here is to learn functions from data. However, these functions are represented as programs that use neural network modules as well as symbolic primitives and are induced using a mix of symbolic search and gradient-based optimization.

In this talk, I will give an elementary introduction to NSP and show how methods in this area have natural applications in accelerating scientific discovery. Specifically, using applications in behavioral neuroscience, I will show that NSP offers natural ways of incorporating prior knowledge into data-driven scientific discovery and interpreting discovered knowledge. I will conclude with a discussion of some of the open technical challenges in NSP in general and NSP-for-science in particular.