Learning Deep Representations by Mutual Information Estimation and Maximization
Abstract
This work investigates unsupervised learning of representations by maximizing mutual information between an input and the output of a deep neural network encoder. Importantly, we show that structure matters: incorporating knowledge about locality in the input into the objective can significantly improve a representation's suitability for downstream tasks. We further control characteristics of the representation by matching to a prior distribution adversarially. Our method, which we call Deep InfoMax (DIM), outperforms a number of popular unsupervised learning methods and compares favorably with fully-supervised learning on several classification tasks in with some standard architectures. DIM opens new avenues for unsupervised learning of representations and is an important step towards flexible formulations of representation learning objectives for specific end-goals.
Cite
Text
Hjelm et al. "Learning Deep Representations by Mutual Information Estimation and Maximization." International Conference on Learning Representations, 2019.Markdown
[Hjelm et al. "Learning Deep Representations by Mutual Information Estimation and Maximization." International Conference on Learning Representations, 2019.](https://mlanthology.org/iclr/2019/hjelm2019iclr-learning/)BibTeX
@inproceedings{hjelm2019iclr-learning,
title = {{Learning Deep Representations by Mutual Information Estimation and Maximization}},
author = {Hjelm, R Devon and Fedorov, Alex and Lavoie-Marchildon, Samuel and Grewal, Karan and Bachman, Phil and Trischler, Adam and Bengio, Yoshua},
booktitle = {International Conference on Learning Representations},
year = {2019},
url = {https://mlanthology.org/iclr/2019/hjelm2019iclr-learning/}
}