We propose a new attribution method for neural networks developed using first principles of causality (to the best of our knowledge, the first such). The neural network architecture is viewed as a Structural Causal Model, and a methodology to compute the causal effect of each feature on the output is presented. With reasonable assumptions on the causal structure of the input data, we propose algorithms to efficiently compute the causal effects, as well as scale the approach to data with large dimensionality. We also show how this method can be used for recurrent neural networks. We report experimental results on both simulated and real datasets showcasing the promise and usefulness of the proposed algorithm.
Aditya Chattopadhyay (Johns Hopkins University)
Piyushi Manupriya (IIT Hyderabad)
Anirban Sarkar (Indian Institute of Technology, Hyderabad)
I'm a 3rd year Ph.D. student at IIT Hyderabad, India. Before that, I completed my bachelors in mathematics and then was in IBM, India for 2.5 years as an associate system engineer. Then I did my masters in computer science before joining Ph.D.
Vineeth N Balasubramanian (Indian Institute of Technology, Hyderabad)
Related Events (a corresponding poster, oral, or spotlight)
2019 Poster: Neural Network Attributions: A Causal Perspective »
Thu Jun 13th 06:30 -- 09:00 PM Room Pacific Ballroom