[Remote event] Introduction to Bayesian deep learning

Bayesian deep learning is an extension of deep learning (DL) using Bayesian statistics. Quantifying uncertainty is the key advantage of incorporating Bayesian tools to DL.

In traditional DL only a single prediction is generated for a given output class. Therefore, there are no mechanisms of quantifying the uncertainty using traditional DL models.

Uncertainty measures are a necessity to better understand a lot of real-world problems. For example, a disease outbreak forecasting model has to incorporate the uncertainty levels in its forecasting to be genuinely useful for healthcare decision makers to use that tool for disease outbreak prevention measures.

A model that generates point predictions is easier to interpret, but a model that generates range of values is more useful to implement as a data driven autonomous decision tool. Bayesian DL therefore ticks a lot of right boxes in making autonomous decision systems more ubiquitous in the real world.

*** What will be covered ***
This is an introductory session on Bayesian DL. We will be covering different approaches to incorporate Bayesian statistics in DL. During the 30 minutes hands-on labs session at the end, we will be using Tensorflow and Tensorflow Probability to build an example Bayesian DL network.

** Agenda **
11am - 11:15am: Introduction and settling down
11:15am - 12noon: Lecture
12noon - 12:30pm: Hands-on lab

** Sign-up sheet **
Please fill the sign-up sheet here: https://bit.ly/moad_20201024
The joining link will be emailed to the participants 24 hour before the event.
(NOTE: Without filling the sign-up sheet, participants will not be allowed to join.)

** Technical requirements **
Use Google Chrome Version 85.x

** Target audience **
Developers, computer programmers or general public with some knowledge of Tensorflow and statistics. All hands-on session contents are using Python3.

*** Related materials ***
A Bayesian deep learning based forecasting tool for COVID19 outbreak in India: https://bit.ly/moad_covid19