Watch a 54-minute lecture from the IPAM's Naturalistic Approaches to Artificial Intelligence Workshop where Professor Levi Lelis from the University of Alberta explores the concept of agents learning behavior libraries for sequential decision-making problems. Discover how artificial agents can develop reusable behavior libraries similar to programming code libraries, enabling them to solve new tasks through behavior composition. Examine practical examples of agents utilizing behavior libraries to accelerate policy learning, including approaches that search program spaces and decompose neural networks into reusable behavioral sub-networks. Gain insights into how agents can continuously learn and store different behaviors while developing the capability to leverage their behavior library for addressing novel challenges.