Generative Adversarial Networks — Hands-on PyTorch Tutorial

Abstract:
This talk will be a hands-on live coding tutorial. We will implement a Generative Adversarial Network (GAN) to learn to generate small images.
GANs are a relatively recent development in unsupervised learning and generative modeling, where we want to learn the distribution of our data. Instead of fitting an explicit density model (with strong assumptions on the data distribution), GANs generate samples, defining an implicit density model. They are able to generate sharp samples from a (meaningful?) continuous latent space.

We will assume only a superficial familiarity with deep learning and a notion of PyTorch. We will make this tutorial as self-contained as possible. The goal is that this talk/tutorial can serve as an introduction to PyTorch at the same time as being an introduction to GANs.

Format —————
The tutorial will be given in Jupyter notebook, fill-in the blank style. Participants are expected to bring laptops, with Jupyter + PyTorch 1.0 already installed (an alternative is to use google Colab).

There will be a short (15min) introduction before we get started with the hands-on implementation.

Event Schedule:

• Doors at 6:15 pm (there will be someone downstairs checking you in)
• Talk begins promptly at 7 pm with Q&A following
• Networking & Drinks!

Food & beverages will be available.

------- Sponsored by Comet.ml ---------
Comet.ml is doing for ML what Github did for software development. We allow data science teams to automagically track their datasets, code changes, experimentation history and production models creating efficiency, transparency, and reproducibility

About our speakers:

Tom Sercu is a Research Engineer at IBM Research AI, working in the IBM T.J. Watson Research Center in Yorktown Heights, NY. He graduated from the MS in Data Science at New York University’s Courant Institute of Mathematical Sciences in May 2015. Before that, he obtained a B.Sc. (2011) and M.Sc. (2013) in Engineering Physics from Ghent University.
His research interests include unsupervised and semi-supervised learning with either no or very small amounts of labeled data, multimodal learning (i.e. learning representations across different data modalities like images, text, and speech), and learning generative models of structured data. He also worked on deep learning approaches to acoustic modeling in speech recognition, bringing advances from the deep learning and computer vision communities to speech recognition. Most recently he worked on Generative Adversarial Networks (GANs), specifically on finding a better distance metric between the data distribution and the generated distribution, which leads to fast and stable training.

Youssef Mroueh is a research staff member in IBM Research AI since April 2015. He received his PhD in computer science in February 2015 from MIT, CSAIL, where he was advised by Professor Tomaso Poggio.
In 2011, he obtained his engineering diploma from Ecole Polytechnique Paris France, and a master of science in Applied Maths from Ecole des Mines de Paris. He is interested in Deep Learning, Machine Learning, Statistical Learning Theory, Computer Vision and Artificial Intelligence. He conducts Modeling and Algorithmic research in Multimodal Deep Learning.