Fast Deep Learning with IPUs & Bayesian Modelling of COVID-19 Data

Dear All.

This is our second Cross-Meetup-Group Virtual Event.

Please register under:

https://register.gotowebinar.com/register/7117739506532791823

AGENDA

OPENING
Jason Ramchandani (Refinitiv): Introduction

TALK 1
Accelerate financial modelling using IPUs and Poplar via standard ML frameworks like TensorFlow
Alexander Tsyplikhin, Senior AI Engineer at Graphcore

In the finance sector, the need for new hardware and software to run complex machine learning models for both training and inference is significant. This talk will outline how Graphcore’s IPU architecture and Poplar® Software Stack powers incredible breakthroughs in Machine Intelligence – and what this means for the future of finance and trading. It will also highlight how to run advanced financial models up to 15x faster using TensorFlow, example use cases and performance benchmarks.

TALK 2
Introduction to Bayesian Modelling using COVID-19 Data
Dr. Thomas Wiecki, VP of Data Science, Head of Research at Quantopian

In this talk, Thomas will demonstrate the benefits of Bayesian statistics using COVID-19 as an example. Specifically, he will show how important uncertainty quantification is, the benefits of hierarchical modelling, and the model development and refinement process, going from a simple exponential model, to a logistic model, to an SIR model.

CLOSING
Closing Remarks

See you all online on Thursday, 14th May 2020!

Yves