Exploratory Gradient Boosting for Reinforcement Learning in Complex Domains

TL;DR

This paper introduces a gradient-boosting function approximator and an exploration strategy to improve reinforcement learning in high-dimensional, complex environments, demonstrating significant performance gains on realistic benchmarks.

Contribution

It presents a novel non-parametric gradient-boosting approach for Q-function approximation and an exploration method inspired by state abstraction, both tailored for complex, high-dimensional RL tasks.

Findings

Outperforms baselines on high-dimensional Minecraft tasks

Maintains competitive performance on standard RL benchmarks

Provides new benchmarks for visual RL environments

Abstract

High-dimensional observations and complex real-world dynamics present major challenges in reinforcement learning for both function approximation and exploration. We address both of these challenges with two complementary techniques: First, we develop a gradient-boosting style, non-parametric function approximator for learning on $Q$-function residuals. And second, we propose an exploration strategy inspired by the principles of state abstraction and information acquisition under uncertainty. We demonstrate the empirical effectiveness of these techniques, first, as a preliminary check, on two standard tasks (Blackjack and $n$-Chain), and then on two much larger and more realistic tasks with high-dimensional observation spaces. Specifically, we introduce two benchmarks built within the game Minecraft where the observations are pixel arrays of the agent's visual field. A combination of our two algorithmic techniques performs competitively on the standard reinforcement-learning tasks while consistently and substantially outperforming baselines on the two tasks with high-dimensional observation spaces. The new function approximator, exploration strategy, and evaluation benchmarks are each of independent interest in the pursuit of reinforcement-learning methods that scale to real-world domains.