Exploiting Language Instructions for Interpretable and Compositional Reinforcement Learning
arXiv:2001.04418
Abstract
In this work, we present an alternative approach to making an agent compositional through the use of a diagnostic classifier. Because of the need for explainable agents in automated decision processes, we attempt to interpret the latent space from an RL agent to identify its current objective in a complex language instruction. Results show that the classification process causes changes in the hidden states which makes them more easily interpretable, but also causes a shift in zero-shot performance to novel instructions. Lastly, we limit the supervisory signal on the classification, and observe a similar but less notable effect.
10 pages, 5 figures
References in corpus (7)
- Human-level performance in first-person multiplayer games with population-based deep reinforcement learning
- Building Generalizable Agents with a Realistic and Rich 3D Environment
- Grounded Language Learning in a Simulated 3D World
- A unified view of entropy-regularized Markov decision processes
- Learning to Understand Goal Specifications by Modelling Reward
- Graying the black box: Understanding DQNs
- Interactive Grounded Language Acquisition and Generalization in a 2D World