Bayesian Deep Learning: Dealing with Uncertainty and Non-Stationarity

Join us on June 27th for an exciting meetup! Quantopian's Director of Data Science, Dr. Thomas Wiecki, will be presenting "Bayesian Deep Learning: Dealing with Uncertainty and Non-Stationarity". As always we will be providing pizza and beer so don't miss out!

Bayesian Deep Learning: Dealing with Uncertainty and Non-Stationarity

Deep Learning continues to build out its dominance over other machine learning approaches on several challenging tasks including image, hand-writing, and speech recognition, image synthesis, as well as playing board and computer games exceeding human expert abilities.

This has generated a lot of interest in the quant finance community to try and mirror Deep Learning's success in the domain of algorithmic trading. Unfortunately, algorithmic trading poses a unique set of challenges. Specifically, the risk (i.e. uncertainty) of certain trading decisions as well as the fact that market behavior changes over time (i.e. non-stationarity) is not handled well by deep learning.

In this talk, I will show how we can embed Deep Learning in the Probabilistic Programming framework PyMC3 and elegantly solve these issues. Expressing neural networks as a Bayesian model naturally instills uncertainty in its predictions. This talk is focused on practitioners and will be introductory and hands-on with many code examples.

About the Speaker

Dr. Thomas Wiecki is Director of Data Science at Quantopian Inc, where he uses Probabilistic Programming and Machine Learning to solve problems in quantitative finance. He has developed various open source projects, such as Pyfolio (https://blog.quantopian.com/pyfolio/) -- a portfolio and risk analysis library, and PyMC3 — a probabilistic programming framework written in Python.

Prior to joining Quantopian, Thomas did his PhD at Brown University where he developed Bayesian methods and Neural Networks to understand brain disorders. A recognized international speaker, he has given talks at conferences across the US, Europe, and Asia.

About Quantopian

Quantopian inspires talented people from around the world to write investment algorithms. They provide capital, data, education tools, and infrastructure to algorithm authors. Quantopian offers license agreements for algorithms that fit its investment strategy, and the licensing authors are paid based on their strategy’s individual performance. Quantopian provides everything a quant needs to create a strategy and profit from it. In the second quarter of this year, Quantopian expects to begin allocating external capital toward these strategies via a pooled investment vehicle.

For more information about Quantopian, please visit: https://www.quantopian.com/ .