Yoshua Bengio

TL;DR: This work proposes an alternative method that is able to recover, in a non-uniform-prior setting, the expressiveness and the desired properties of the Laplacian representation, and shows that a simple augmentation of the representation objective with the learned temporal abstractions improves dynamics-awareness and helps exploration.

TL;DR: Experimental results support the hypothesis that modularity is a robust prior for learning invariant relational reasoning, and show that very shallow ResMixNet models are capable of learning each of the two tasks well, attaining less than 2% and 1% test error on the MNIST Parity and the colorized Pentomino tasks respectively.

TL;DR: The factorized embeddings (FE) model is presented, a self-supervised deep learning algorithm that learns simultaneously, by tensor factorization, gene and sample representation spaces, and it is found that the sample representation captures information on single gene and global gene expression patterns.

TL;DR: In this paper, an adversarial domain adaptation network is used to match the empirical probability distribution of the residuals of the reconstructed neural signals. But, the authors do not address the domain adaptation problem in this paper.

TL;DR: In this paper, the authors proposed a method that learns and optimizes (during training) a source-dependent mask and does not need the aforementioned post processing step, which shows an increase of 0.49 dB for the signal to distortion ratio and 0.30 dB for signal to interference ratio, compared to previous state-of-the-art approaches for monaural singing voice separation.