I am an applied researcher. My research area is "explanation", i.e., I build theories and tools to explain machine learning model predictions and separately, changes in business metrics. I have also worked on, and published in the areas of question-answering systems, ad auctions, security protocol analysis, privacy, and computational biology.
(This is my only presence on the internet.)
Deep Learning Explanations
- Integrated Gradients (paper@ICML17, code): Feature attributions for deep networks
- Conductance (paper@ICLR19): Feature attribution for neurons
- paper@ACL18: Constructing attacks using feature attributions
- Visualizing attributions (paper): principled visualization of feature attributions for Image networks
- A critique about a critique of our work.