Combating False Negatives in Adversarial Imitation Learning (Student Abstract)

Konrad Zolna
Konrad Zolna
Chitwan Saharia
Chitwan Saharia
Léonard Boussioux
Léonard Boussioux
David Yu-Tung Hui
David Yu-Tung Hui
Maxime Chevalier-Boisvert
Maxime Chevalier-Boisvert

AAAI, pp. 13999-14000, 2020.

Cited by: 0|Bibtex|Views79|DOI:https://doi.org/10.1609/AAAI.V34I10.7272
EI
Other Links: dblp.uni-trier.de|academic.microsoft.com

Abstract:

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 sampl...More

Code:

Data:

Your rating :
0

 

Tags
Comments