Titre : Generative Adversarial Networks and Few-shot Learning
Lieu : Local 3195, Pavillon André-Aisenstadt, Université de Montréal 
Date : Mercredi 12 décembre à 13h
Jury : 
Président : Emma Frejinger
Membre : Liam Paull
Directeur : Yoshua Bengio

Résumé:

GANs are a class of state-of-the-art generative models which provides significant improvements on many deep learning tasks. For example, image translation, domain adaptation and semi-supervised learning. In this Predoc report, we make a review of GAN models and discuss specifically a problem of GAN models called mode dropping. Then we provide a conceptually simple and general framework called MetaGAN for few-shot learning problems. Most state-of-the-art few-shot classification models can be integrated with MetaGAN in a principled and straightforward way. By introducing an adversarial generator conditioned on tasks, we augment vanilla few-shot classification models with the ability to discriminate between real and fake data. We argue that this GAN-based approach can help few-shot classifiers to learn sharper decision boundary, which could generalize better. We show that with our MetaGAN framework, we can extend supervised few-shot learning models to naturally cope with unlabeled data. Different from previous work in semi-supervised few-shot learning, our algorithms can deal with semi-supervision at both sample-level and task-level. We give theoretical justifications of the strength of MetaGAN, and validate the effectiveness of MetaGAN on challenging few-shot image classification benchmarks.

 
Vous êtes cordialement invité.