Timezone: »
We propose an approach to self-supervised representation learning based on maximizing mutual information between features extracted from multiple views of a shared context. For example, one could produce multiple views of a local spatio-temporal context by observing it from different locations (e.g., camera positions within a scene), and via different modalities (e.g., tactile, auditory, or visual). Or, an ImageNet image could provide a context from which one produces multiple views by repeatedly applying data augmentation. Maximizing mutual information between features extracted from these views requires capturing information about high-level factors whose influence spans multiple views – e.g., presence of certain objects or occurrence of certain events. Following our proposed approach, we develop a model which learns image representations that significantly outperform prior methods on the tasks we consider. Most notably, using self-supervised learning, our model learns representations which achieve 68.1% accuracy on ImageNet using standard linear evaluation. This beats prior results by over 12% and concurrent results by 7%. When we extend our model to use mixture-based representations, segmentation behaviour emerges as a natural side-effect. Our code is available online: https://github.com/Philip-Bachman/amdim-public.
Author Information
Philip Bachman (Microsoft Research)
R Devon Hjelm (Microsoft Research)
William Buchwalter (Microsoft)
More from the Same Authors
-
2020 Poster: Deep Reinforcement and InfoMax Learning »
Bogdan Mazoure · Remi Tachet des Combes · Thang Long Doan · Philip Bachman · R Devon Hjelm -
2019 Poster: Unsupervised State Representation Learning in Atari »
Ankesh Anand · Evan Racah · Sherjil Ozair · Yoshua Bengio · Marc-Alexandre Côté · R Devon Hjelm -
2019 Poster: On Adversarial Mixup Resynthesis »
Christopher Beckham · Sina Honari · Alex Lamb · Vikas Verma · Farnoosh Ghadiri · R Devon Hjelm · Yoshua Bengio · Chris Pal -
2016 Poster: An Architecture for Deep, Hierarchical Generative Models »
Philip Bachman -
2015 Poster: Data Generation as Sequential Decision Making »
Philip Bachman · Doina Precup -
2015 Spotlight: Data Generation as Sequential Decision Making »
Philip Bachman · Doina Precup -
2014 Poster: Learning with Pseudo-Ensembles »
Philip Bachman · Ouais Alsharif · Doina Precup