Causality for Machine Learning

Machine learning allows us to detect subtle correlations in large data sets, and use those correlations to make accurate predictions. However, these subtle correlations are often spurious - they exist only in a particular dataset - and the resultant model performs poorly, or gives unexpected results in the real world. Moreover, reasoning based on spurious correlations is dangerous. Business decisions should be based on things that are true, not things that are true only in a limited dataset. The trouble, of course, is identifying what is spurious and what is not. In this webinar, we’ll explain how combining causal inference with machine learning can help us address these problems. WE’LL COVER: when you should think about causality and lessons to apply in your data science practice the latest research at the intersection of machine learning and causality how causal thinking helps us write models that generalize to new circumstances, including an example of the causal approach applied to a computer vision problem