Y
Yoshua Bengio
Researcher at Université de Montréal
Publications - 1146
Citations - 534376
Yoshua Bengio is an academic researcher from Université de Montréal. The author has contributed to research in topics: Artificial neural network & Deep learning. The author has an hindex of 202, co-authored 1033 publications receiving 420313 citations. Previous affiliations of Yoshua Bengio include McGill University & Centre de Recherches Mathématiques.
Papers
More filters
Journal ArticleDOI
Towards Non-Saturating Recurrent Units for Modelling Long-Term Dependencies
TL;DR: This work proposes a new recurrent architecture (Non-saturating Recurrent Unit; NRU) that relies on a memory mechanism but forgoes both saturating activation functions and saturating gates, in order to further alleviate vanishing gradients.
Posted Content
Wasserstein Dependency Measure for Representation Learning
TL;DR: In this article, the Wasserstein distance is used instead of the KL divergence in the mutual information estimator to learn more complete representations for unsupervised representation learning, and a practical approximation to this theoretically motivated solution is constructed using Lipschitz constraint techniques from the GAN literature.
Proceedings Article
On the challenges of physical implementations of RBMs
TL;DR: In this paper, the authors conduct software simulations to determine how harmful each of these restrictions is, and suggest that designers of new physical computing hardware and algorithms for physical computers should focus their efforts on overcoming the limitations imposed by the topology restrictions of currently existing physical computers.
Proceedings Article
Variational Walkback: Learning a Transition Operator as a Stochastic Recurrent Net
TL;DR: In this article, the authors proposed a variational walkback (VW) method to learn a stochastic transition operator whose repeated application provides generated samples, which can be used to learn more biologically plausible asymmetric neural networks and more general non-energy based dynamical systems.
Journal ArticleDOI
On the search for new learning rules for ANNs
TL;DR: A framework where a learning rule can be optimized within a parametric learning rule space and a theoretical study of their generalization properties when estimated from a set of learning tasks and tested over another set of tasks is presented.