#21 Gaussian Processes, Bayesian Neural Nets & SIR Models, with Elizaveta Semenova

Aug 13, 2020

Elizaveta Semenova, a postdoctorate in Bayesian Machine Learning, discusses her work on Gaussian Processes for studying the spread of Malaria and fitting dose-response curves in pharmaceutical tests. She also talks about her latest paper on Bayesian neural networks for drug toxicity prediction and the interesting link between BNNs and Gaussian Processes.

Ask episode

AI Snips

Chapters

Transcript

Episode notes

ANECDOTE

From Math to Epidemiology

Elizaveta Semenova initially pursued theoretical mathematics but found it too abstract.
She transitioned to epidemiology, driven by a passion for global health and a desire for more applied work.

INSIGHT

Bayesian First

Elizaveta Semenova's first exposure to statistical modeling was through Bayesian methods during her PhD.
She was later surprised to discover the prevalence of non-Bayesian approaches in the field.

ADVICE

Slow but Exact

Prioritize slow but exact MCMC solutions over faster approximate methods when time isn't critical.
A better posterior distribution is worth the computational cost in many research contexts.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

I bet you heard a lot about epidemiological compartmental models such as SIR in the last few months? But what are they exactly? And why are they so useful for epidemiological modeling?

Elizaveta Semenova will tell you why in this episode, by walking us through the case study she recently wrote with the Stan team. She’ll also tell us how she used Gaussian Processes on spatio-temporal data, to study the spread of Malaria, or to fit dose-response curves in pharmaceutical tests.

And finally, she’ll tell us how she used Bayesian neural networks for drug toxicity prediction in her latest paper, and how Bayesian neural nets behave compared to classical neural nets. Ow, and you’ll also learn an interesting link between BNNs and Gaussian Processes…

I know: Liza works on _a lot_ of projects! But who is she? Well, she’s a postdoctorate in Bayesian Machine Learning at the pharmaceutical company AstraZeneca, in Cambridge, UK.

Elizaveta did her masters in theoretical mathematics in Moscow, Russia, and then worked in financial services as an actuary in various European countries. She then did a PhD in epidemiology at the University of Basel, Switzerland. This is where she got interested in health applications – be it epidemiology, global health or more small-scale biological questions. But she’ll tell you all that in the episode ;)

Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his awesome work at https://bababrinkman.com/ !

Links from the show:

Liza on Twitter: https://twitter.com/liza_p_semenova
Liza on GitHub: https://github.com/elizavetasemenova
Liza's blog: https://elizavetasemenova.github.io/blog/
A Bayesian neural network for toxicity prediction: https://www.biorxiv.org/content/10.1101/2020.04.28.065532v2
Bayesian Neural Networks for toxicity prediction -- Video presentation: https://www.youtube.com/watch?v=BCQ2oVlu_tY&t=751s
Bayesian workflow for disease transmission modeling in Stan: https://mc-stan.org/users/documentation/case-studies/boarding_school_case_study.html
Andrew Gelman's comments on the SIR case-study: https://statmodeling.stat.columbia.edu/2020/06/02/this-ones-important-bayesian-workflow-for-disease-transmission-modeling-in-stan/
Determining organ weight toxicity with Bayesian causal models: https://www.biorxiv.org/content/10.1101/754853v1
Material for Applied Machine Learning Days ("Embracing uncertainty"): https://github.com/elizavetasemenova/EmbracingUncertainty
Predicting Drug-Induced Liver Injury with Bayesian Machine Learning: https://pubs.acs.org/doi/abs/10.1021/acs.chemrestox.9b00264
Ordered Logistic Regression in Stan, PyMC3 and Turing: https://medium.com/@liza_p_semenova/ordered-logistic-regression-and-probabilistic-programming-502d8235ad3f
PyMCon website: https://pymc-devs.github.io/pymcon/
PyMCon Call For Proposal: https://pymc-devs.github.io/pymcon/cfp
PyMCon Sponsorship Form: https://docs.google.com/forms/d/e/1FAIpQLSdRDI1z0U0ZztONOFiZt2VdsBIZtAWB4JAUA415Iw8RYqNbXQ/viewform
PyMCon Volunteer Form: https://docs.google.com/forms/d/e/1FAIpQLScCLW5RkNtBz1u376xwelSsNpyWImFisSMjZGP35fYi2QHHXw/viewform

Thank you to my Patrons for making this episode possible!

Yusuke Saito, Avi Bryant, Ero Carrera, Brian Huey, Giuliano Cruz, Tim Gasser, James Wade, Tradd Salvo, Adam Bartonicek, William Benton, Alan O'Donnell, Mark Ormsby, Demetri Pananos, James Ahloy, Jon Berezowski, Robin Taylor, Thomas Wiecki, Chad Scherrer, Vincent Arel-Bundock, Nathaniel Neitzke, Zwelithini Tunyiswa, Elea McDonnell Feit, Bertrand Wilden, James Thompson, Stephen Oates, Gian Luca Di Tanna, Jack Wells, Matthew Maldonado, Ian Costley, Ally Salim, Larry Gill, Joshua Duncan, Ian Moran and Paul Oreto.