Learning, Reasoning, and Planning with Neuro-Symbolic Concepts

Join this colloquium talk where Jiayuan Mao from MIT presents "Learning, Reasoning, and Planning with Neuro-Symbolic Concepts." Discover a framework for building intelligent agents that can continually learn, reason, and plan using neuro-symbolic concepts - compositional abstractions of the physical world that represent object properties, relations, and actions. Learn how these concepts combine symbolic programs with modular neural networks to achieve superior data efficiency, faster reasoning and planning, and strong generalization capabilities. See demonstrations across visual reasoning in 2D/3D environments, motion analysis, video data, and decision-making tasks in both virtual and robotic manipulation contexts. Mao, a PhD student advised by Josh Tenenbaum and Leslie Kaelbling, has received multiple Best Paper awards and was named a Rising Star in EECS and Generative AI (2024).