ML Anthology

Combating False Negatives in Adversarial Imitation Learning (Student Abstract)

Konrad Zolna, Chitwan Saharia, Léonard Boussioux, David Yu-Tung Hui, Maxime Chevalier-Boisvert, Dzmitry Bahdanau, Yoshua Bengio
AAAI 2020 pp. 13999-14000
doi:10.1609/AAAI.V34I10.7272 /aaai/2020/zolna2020aaai-combating/

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 sample efficiency over the baselines by at least an order of magnitude.

PDF AAAI Semantic Scholar

Cite

Text

Zolna et al. "Combating False Negatives in Adversarial Imitation Learning (Student Abstract)." AAAI Conference on Artificial Intelligence, 2020. doi:10.1609/AAAI.V34I10.7272

Markdown

[Zolna et al. "Combating False Negatives in Adversarial Imitation Learning (Student Abstract)." AAAI Conference on Artificial Intelligence, 2020.](https://mlanthology.org/aaai/2020/zolna2020aaai-combating/) doi:10.1609/AAAI.V34I10.7272

BibTeX

@inproceedings{zolna2020aaai-combating,
  title     = {{Combating False Negatives in Adversarial Imitation Learning (Student Abstract)}},
  author    = {Zolna, Konrad and Saharia, Chitwan and Boussioux, Léonard and Hui, David Yu-Tung and Chevalier-Boisvert, Maxime and Bahdanau, Dzmitry and Bengio, Yoshua},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {2020},
  pages     = {13999-14000},
  doi       = {10.1609/AAAI.V34I10.7272},
  url       = {https://mlanthology.org/aaai/2020/zolna2020aaai-combating/}
}