Online variance-reducing optimization

Nicolas Le Roux, Reza Babanezhad, Pierre-Antoine Manzagol

Feb 12, 2018 (modified: Feb 12, 2018) ICLR 2018 Workshop Submission readers: everyone
  • Abstract: We emphasize the importance of variance reduction in stochastic methods and propose a probabilistic interpretation as a way to store information about past gradients. The resulting algorithm is very similar to the momentum method, with the difference that the weight over past gradients depends on the distance moved in parameter space rather than the number of steps.
  • TL;DR: Online variance reduction looks a lot like momentum
  • Keywords: online learning, momentum