Dartmouth Events

Deep Learning is a fast developing topic in machine learning, aiming at designing models able to automatically extract rich, multi-layer representations of data. It is motivated by its potential to avoid the hard work of manually defining features for each type of data at hand. Deep learning has demonstrated its ability to sidestep this previously ubiquitous step of machine learning on a large variety of artificial intelligence tasks, in computer vision, natural language processing and speech recognition .

In this talk, I will argue that the real challenge ahead lies in the exploitation of unlabelled data, which is now overwhelmingly abundant in this era of Big Data. I will describe my recent work in this direction. This work thus relies on the development of new unsupervised models, that better scale to the context of Big Data, both computationally and statistically.

Specifically, I will present the application of such models to the modelling of topics in text data, to the classification of multilingual documents and to multimodal classification of pairs of images and texts.

Hugo Larochelle is Assistant Professor at the Université de Sherbrooke (UdeS). Before joining the Computer Science department of UdeS in 2011, he spent two years in the machine learning group at University of Toronto, as a postdoctoral fellow under the supervision of Geoffrey Hinton. He obtained his Ph.D. at Université de Montréal, under the supervision of Yoshua Bengio. He is the recipient of two Google Faculty Awards, acts as associate editor for the IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) and is a member of the editorial board of the Journal of Artificial Intelligence Research (JAIR).