Combating False Negatives in Adversarial Imitation Learning (Student Abstract)

2020-06-02

Proceedings of the AAAI Conference on Artificial Intelligence, 34

Volume

Vol. 34 No. 10: Issue 10: AAAI-20 Student Tracks

Student Abstract Track

We define the False Negatives problem and show that it is a significant limitation in adversarial imitation learning. We propose a method that solves the problem by leveraging the nature of goal-conditioned tasks. The method, dubbed Fake Conditioning, is tested on instruction following tasks in BabyAI environments, where it improves sample efficiency over the baselines by at least an order of magnitude.

10.1609/aaai.v34i10.7272