# Growing Action Spaces

## TL;DR

This paper introduces a curriculum-based method for progressively enlarging action spaces in reinforcement learning, significantly improving learning efficiency in complex, large-scale tasks like StarCraft micromanagement.

## Contribution

It proposes a novel approach that uses off-policy RL to transfer knowledge across growing action spaces, enabling faster learning in complex environments.

## Key findings

- Effective in control tasks with small action spaces
- Improves learning speed in large-scale StarCraft tasks
- Demonstrates successful transfer of knowledge across action spaces

## Abstract

In complex tasks, such as those with large combinatorial action spaces, random exploration may be too inefficient to achieve meaningful learning progress. In this work, we use a curriculum of progressively growing action spaces to accelerate learning. We assume the environment is out of our control, but that the agent may set an internal curriculum by initially restricting its action space. Our approach uses off-policy reinforcement learning to estimate optimal value functions for multiple action spaces simultaneously and efficiently transfers data, value estimates, and state representations from restricted action spaces to the full task. We show the efficacy of our approach in proof-of-concept control tasks and on challenging large-scale StarCraft micromanagement tasks with large, multi-agent action spaces.