Probabilistic Programming using Edward/TensorFlow

** The Material on this site and github would be updated in following months before and during the conference.

Abstract

Researchers have built probabilistic models and deep learning models that have provided benefits in various domains. The probabilistic programming platforms and languages of today empower non-experts to create and apply such probabilistic models based on Bayesian inference techniques. In this tutorial, we combine the power of Edward and TensorFlow to teach how to apply probabilistic programming and deep learning for use cases such as dimensionality reduction and classification in computer vision and image processing.

Duration: Half day (3 hours, 30 minutes, plus 30 minutes break)

Tutorial Objectives

• The audience will gain knowledge of the probabilistic programming.
• The tutorial will use TensorFlow Probability and Edward2 for two hands on examples.

Tutorial Outline ( under revision)

The tutorial will cover the following sessions of 30 minutes each:

Session 1: Introduction to Probabilistic Programming - Theory

• What is Probabilistic Programming
• Probabilistic Machine Learning
• Variational Inference

Session 2: Probabilistic Programming Tools

• Bugs, Stan, Pyro
• Tensorflow Probability and Edward 2

Session 3: Variational Autoencoders - Practical Hands-On

Session 4: Bayesian Neural Network - Practical Hands-On

Session 5: Road Ahead: Further Learning and Research Resources

Prerequisite knowledge:

The audience should be aware of the basic probability, bayesian inference and deep learning algorithms. Basic knowledge of Python language and one of the deep learning frameworks such as PyTorch or TensorFlow will be useful.

Supplementary materials:

• Neural Networks Basics
• Convolutional Neural Networks
• TensorFlow

Presenters

Dr. Amita Kapoor, Associate Professor, University of Delhi, dr.amita.kapoor@ieee.org

Amita Kapoor is Associate Professor in the Department of Electronics, SRCASW, University of Delhi. She has been actively teaching neural networks for the last twenty years. She is the recipient of the prestigious DAAD fellowship to pursue a part of her research work in Karlsruhe Institute of Technology, Karlsruhe, Germany. She had been awarded Best Presentation Award at International Conference Photonics 2008 for her paper. She has more than 40 research publications in international journals and conferences. Her present research areas include Machine Learning, Artificial Intelligence, Neural Networks, Robotics, Buddhism (Philosophy and Psychology) and Ethics in AI. She is also the author of the book Tensorflow 1.x Deep Learning Cookbook by Packt Publishers.

Armando Fandango, Affiliation To Be Announced

Armando creates AI empowered products by leveraging reinforcement learning, deep learning, machine learning, distributed computing, and computational methods. Armando has provided thought leadership as Chief Data Scientist and Director at small and large organizations. Armando has been advising high-tech AI-based startups. Armando has authored books titled Python Data Analysis - Second Edition and Mastering TensorFlow. He has published research in international journals and conferences.