Self-supervised representation learning using bootstrapped latent representations

Jean-Bastien Grill, Florian Strub, Florent Altché, Corentin Tallec, Pierre Richemond, Bernardo Ávila Pires, Zhaohan Daniel Guo, Mohammad Gheshlaghi Azar, Bilal Piot, Rémi Munos, Michal Valko

2021 · US Patent App. 17/338,777 (2021)

Abstract

Training a target encoder neural network to generate representations of images using a self-supervised approach based on bootstrapping latent representations, without requiring negative pairs or specialized architectures.

PDF · US Patent App. 17/338,777