Explore projects

Experiments to see if removing the part of the gradient that lies in the dominating eigenspace for hessian can make the optimization result better off.

Measurements of Three-Level Hierarchical Structure in the Outliers in the Spectrum of Deepnet Hessians (ICML 2019)

Optimizing PAC-Bayes bounds for Stochastic Neural Networks with Gaussian weights

Framework for perturbation analysis on Spark. Used for computational fact-checking/finding/discovery tasks.

Short-graph Fourier transform