Work

A Deep Hierarchical Approach to Lifelong Learning in Minecraft

Year: 2017

Type: conference-paper

Abstract: We propose a lifelong learning system that has the ability to reuse and transfer knowledge from one task to another while efficiently retaining the previously learned knowledge-base. Knowledge is transferred by learning reusable skills to solve tasks in Minecraft, a popular video game which is an unsolved and high-dimensional lifelong learning problem. These reusable skills, which we refer to as Deep Skill Networks, are then incorporated into our novel Hierarchical Deep Reinforcement Learning Network (H-DRLN) architecture using two techniques: (1) a deep skill array and (2) skill distillation, our novel variation of policy distillation (Rusu et. al. 2015) for learning skills. Skill distillation enables the H-DRLN to efficiently retain knowledge and therefore scale in lifelong learning, by accumulating knowledge and encapsulating multiple reusable skills into a single distilled network. The H-DRLN exhibits superior performance and lower learning sample complexity compared to the regular Deep Q Network (Mnih et. al. 2015) in sub-domains of Minecraft.

Source: Proceedings of the AAAI Conference on Artificial Intelligence

DOI: 10.1609/aaai.v31i1.10744

Authors: Chen Tessler, Shahar Givony, Tom Zahavy, Daniel Mankowitz, Shie Mannor

Institution: Technion – Israel Institute of Technology

Language: English

Topic: Reinforcement Learning in Robotics

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: diamond

Funders: Intel Collaboration Research Institute for Computational Intelligence, FP7 Ideas: European Research Council

Award: Scaling Up Reinforcement Learning: Structure Learning, Skill Acquisition, and Reward Shaping""