Software

Software package developed for the paper Training Set Debugging using Trusted Items (AAAI 2018). DUTI finds, for general ERM type learners, the most plausible flaws in training data that results in particular prediction errors at test time. Code available here.

InterpretML is an open-source python package for training interpretable machine learning models and explaining blackbox systems, developped as part of my summer 2018 internship at Microsoft Research with Rich Caruana. [Gihub repo].