Planning with Arithmetic and Geometric Attributes

TL;DR

This paper demonstrates that incorporating geometric and arithmetic structures into environment attributes significantly improves an agent's ability to generalize and learn efficiently in structured environments.

Contribution

It introduces a method to embed geometric and arithmetic structures into environment attributes, enhancing sample efficiency and generalization in structured tasks.

Findings

Enhanced sample complexity with structured attributes

Faster generalization to novel tasks

Improved task composition capabilities

Abstract

A desirable property of an intelligent agent is its ability to understand its environment to quickly generalize to novel tasks and compose simpler tasks into more complex ones. If the environment has geometric or arithmetic structure, the agent should exploit these for faster generalization. Building on recent work that augments the environment with user-specified attributes, we show that further equipping these attributes with the appropriate geometric and arithmetic structure brings substantial gains in sample complexity.