Automated curriculum generation for Policy Gradients from Demonstrations
This addresses sample efficiency in RL for instruction following, but is incremental as it builds on existing curriculum learning and demonstration-based approaches.
The paper tackles the problem of training reinforcement learning agents for instruction following by developing an automated curriculum generation method that uses expert demonstrations and trains agents from goal to start. The method shows improved sample efficiency on some BabyAI tasks compared to a PPO baseline.
In this paper, we present a technique that improves the process of training an agent (using RL) for instruction following. We develop a training curriculum that uses a nominal number of expert demonstrations and trains the agent in a manner that draws parallels from one of the ways in which humans learn to perform complex tasks, i.e by starting from the goal and working backwards. We test our method on the BabyAI platform and show an improvement in sample efficiency for some of its tasks compared to a PPO (proximal policy optimization) baseline.