Fast Inference and Transfer of Compositional Task Structures for Few-shot Task Generalization

TL;DR

This paper introduces a method for few-shot reinforcement learning that infers common subtask graph structures to enable faster adaptation to new complex tasks, outperforming existing algorithms.

Contribution

The paper proposes a multi-task subtask graph inferencer that captures shared high-level task structures to improve task inference and generalization in few-shot settings.

Findings

Faster adaptation to unseen tasks compared to existing methods

Effective inference of subtask graph structures from training tasks

Improved performance in 2D grid-world and web navigation domains

Abstract

We tackle real-world problems with complex structures beyond the pixel-based game or simulator. We formulate it as a few-shot reinforcement learning problem where a task is characterized by a subtask graph that defines a set of subtasks and their dependencies that are unknown to the agent. Different from the previous meta-rl methods trying to directly infer the unstructured task embedding, our multi-task subtask graph inferencer (MTSGI) first infers the common high-level task structure in terms of the subtask graph from the training tasks, and use it as a prior to improve the task inference in testing. Our experiment results on 2D grid-world and complex web navigation domains show that the proposed method can learn and leverage the common underlying structure of the tasks for faster adaptation to the unseen tasks than various existing algorithms such as meta reinforcement learning, hierarchical reinforcement learning, and other heuristic agents.