11-04, 13:00–13:30 (Asia/Jerusalem), Green Track
Imagine the newest medical prediction algorithm is claiming you have high-risk for some health condition. I bet the first thing going through your mind is "well, what can I do to reduce it".
Regular prediction is not always enough, we often care about predicting the consequences of several paths of action we can take - the causal effect of these actions.
In this talk I will briefly present causal inference - the science of estimating causal effect of actions using observational data and how it differs from regular prediction. I will overview models for estimating causal effect and how to apply them with causallib - a one-stop-shop open-source Python package for flexible causal inference modeling.
Causal inference is the science of estimating the causal effects of actions using non-experimental data.
In this talk I will introduce causal inference, how it differs from the more familiar machine learning predictions, and why it is a harder task. I will present an overview of statistical models that can estimate causal effects, and I will present causallib - an open-source one-stop-shop Python package for flexible causal inference I created and maintain.
The main objective of the talk is to familiarize participants with the field of causal inference, increasing their awareness of the limitations in more common prediction models. A secondary objective is to present the tools that may help obtain causal inferences using a package which design corresponds with the scientific ecosystem in Python.
The talk is mainly aimed for data scientists familiar with machine learning, but group leaders may also benefit from understanding the limitations of regular prediction models and that they may be overcome.
Ehud is a research staff member at IBM Research, marrying machine learning with causal inference to address questions in medicine and healthcare.
He combines applied research with tool development for research, having created and currently maintaining Causallib—an open-source Python package for flexible causal inference modeling—used by many practitioners in both academia and industry. Over his 8 years at IBM, he has led the causality strategy for the company's global efforts in drug discovery, consulted to many of its research labs worldwide, lectured on causality to staff and clients, developed novel methodologies and published his research.
He holds an MSc. in computer science and computational biology from the Hebrew University, where he worked on trait prediction using DNA and assessed its potential consequences for population genetics and embryo selection. A musician and hiker, but mostly a parent.