Explore time-continuous neural networks and Liquid Time Constant (LTC) networks in this 31-minute talk by Ramin Hasani from MIT. Delve into the concept of stable state and time-constant in LTCs, examine their expressivity, and understand their performance characteristics. Learn about Neural Circuit Policies as Dynamic Causal Models and gain insights into the synthesis of models and systems in the field of artificial intelligence.
Overview
Syllabus
Intro
What is a time-continuous neural network?
Time-continuous neural networks
LTC have stable state and time-constant
Expressivity
Summary
LTCs: Performance
Neural Circuit Policies are Dynamic Causal Models
Taught by
Simons Institute
