In few-shot classification, we are interested in learning algorithms that train a classifier from only a handful of labeled examples. Recent progress in few-shot classification has featured meta-learning, in which a parameterized model for a learning algorithm is defined and trained on episodes representing different classification problems, each with a small labeled training set and its corresponding test set. In this work, we advance this few-shot classification paradigm towards a scenario where unlabeled examples are also available within each episode. To address this paradigm, we propose novel extensions of Prototypical Networks (Snell et al., 2017) that are augmented with the ability to use unlabeled examples when producing prototypes. Our experiments confirm that our Prototypical Networks can learn to improve their predictions due to unlabeled examples, much like a semi-supervised algorithm would.

Bio: Hugo Larochelle is a research scientist and lead of the Google Brain team in Montreal. He is also an adjunct professor at the Université de Montréal and at the Université de Sherbrooke, as well as Associate Director of the Learning in Machines and Brains program of the Canadian Institute for Advanced Research.

Note: The presentation will be given in French (the slides will be in English). The seminar will be presented in room PLT-2700 at 10:30 a.m.