Object Representations for Learning and Reasoning

Thirty-fourth Conference on Neural Information Processing Systems (NeurIPS)

December 11, 2020, Virtual Workshop

Join via the livestream 🎥 · RocketChat · @ORLR_Workshop · #ORLR2020 · Join our community Slack!

---

Abstract

We all seem to know intuitively what an object is, however the notion of an object appears to be very hard to define precisely. In order to move forward in building better object representations for learning and reasoning, it is important to first agree on what objects are. In this talk I will propose an operational definition of objects using insights from physics. I will then use this definition to suggest what a good object-centered representation should look like, and why it might be useful for more efficient and robust learning and reasoning.