Sumedha Singla (Ph.D. student)
University of Pittsburgh
Using Causal Analysis for Interpretability
This talk by Sumedha explains their latest work about explaining black-boxes such as Deep Neural Networks in a smooth way by using counterfactuals and causal analysis.
!! Soon we will post an additional video where Sumedha explains another of her works: Using Causal Analysis for Conceptual Explanations of Deep Learning
The works presented in the talk are the following:
1. Singla S, Pollack B, Chen J, Batmanghelich K. Explanation by Progressive Exaggeration. In International Conference on Learning Representations 2019 Sep 25
2. Singla S, Pollack B, Wallace S, Batmanghelich K. Explaining the black-box smoothly – A counterfactual approach. Under review.
Extra Readings:
- Peters, Jonas, Dominik Janzing, and Bernhard Schölkopf. Elements of causal inference: foundations and learning algorithms. The MIT Press, 2017.
There are no assignments for this class.
Introduction to AI Interpretability 