Discover more from Altdeep.ai Newsletter
Webinar: Causal Machine Learning Blitz
A 60 minute blitz on causal machine learning with code examples
I am pleased to invite you to attend a free webinar I am hosting entitled Causal Machine Learning Blitz. It runs this Thursday, January 9, at 11AM Eastern time.
What you’ll get from the webinar
Quickly get up to speed on how tech companies like Netflix and Facebook make decisions with causal inference, and how causality fits into cutting-edge AI research.
Learn the 4 essential R and Python causal inference libraries for reasoning on directed graphs and identifying/estimating causal effects.
Get code examples for interventional and counterfactual reasoning with Pyro, a PyTorch-based deep learning library.
Get code examples for counterfactual reasoning on simulation programs, featuring ODE-modeling examples in Stan.
Get a chance to ask questions during a live Q&A session.
Sure, causality matters, but so what?
One of my big take-aways from the machine learning conference NeurIPS in December was that causal inference is having a moment in the machine learning community.
I also learned that the “so-what” is missing. One the one hand, it is clear to everyone that state-of-the-art deep learning struggles with causality.
For example, in an online interface to a cutting-edge deep learning-based language model (called a transformer network), I entered
She was a vegan, and so she ordered lentils. Had she not been vegan, she might have ordered _________
Basic causal reasoning tells you the answer should be “meat” or something meat-related. Instead, the answer was as follows.
OK, but so what? We know building AI that can reason causally is hard, but what do we do about it? Where do we even start?
On the other hand, when the machine learning community looks to causal inference researchers and practitioners, they tend to get explanations that go deep on specific problems and specific statistical methods for solving those problems. “OK, you taught me how to use instrumental variables to estimate a causal effect in a clinical trial with a certain design, but so what?” How does this generalize to problems I actually care about? What can I do with the tools and techniques I already know?
This webinar solves this problem. It firmly grounds causal modeling on a foundation of generative machine learning, and uses machine learning tools such as PyTorch to show what these models can do and how to reason about them.
After the course, you’ll have concrete ideas and code examples for next steps for applying causal modeling in your machine learning journey.
More of this to come…
The webinar is heralding a set of online courses I am building that gets at questions just like these. Their goal is to bridge the intersection of cutting-edge research and best practices in machine learning engineering. Please stay tuned!