Look where you look! Saliency-guided Q-networks for generalization in visual Reinforcement Learning

1 code implementation • 16 Sep 2022 • David Bertoin, Adil Zouitine, Mehdi Zouitine, Emmanuel Rachelson

This implies that training of a policy with small generalization gap should focus on such important pixels and ignore the others.

reinforcement-learning Reinforcement Learning (RL)

Local Feature Swapping for Generalization in Reinforcement Learning

no code implementations • ICLR 2022 • David Bertoin, Emmanuel Rachelson

We also demonstrate the effectiveness of CLOP as a general regularization technique in supervised learning.

Memorization reinforcement-learning +1

Disentanglement by Cyclic Reconstruction

1 code implementation • 24 Dec 2021 • David Bertoin, Emmanuel Rachelson

This enables the isolation of task-specific information from both domains and a projection into a common representation.

Disentanglement Information Retrieval +2

Numerical influence of ReLU'(0) on backpropagation

1 code implementation • NeurIPS 2021 • David Bertoin, Jérôme Bolte, Sébastien Gerchinovitz, Edouard Pauwels

In theory, the choice of ReLU(0) in [0, 1] for a neural network has a negligible influence both on backpropagation and training.

Numerical influence of ReLU’(0) on backpropagation

no code implementations • NeurIPS 2021 • David Bertoin, Jerome Bolte, Sébastien Gerchinovitz, Edouard Pauwels

In theory, the choice of ReLU(0) in [0, 1] for a neural network has a negligible influence both on backpropagation and training.

Disentangled cyclic reconstruction for domain adaptation

no code implementations • 1 Jan 2021 • David Bertoin, Emmanuel Rachelson

The domain adaptation problem involves learning a unique classification or regres-sion model capable of performing on both a source and a target domain.

Disentanglement Unsupervised Domain Adaptation