DEMO-Net: Degree-specific Graph Neural Networks for Node and Graph Classification

Large Scale benchmarking of state of the art text vectorizers

Slides for my talk at PyData London 2019

FEATure HashER

Sample of Feature Hashing for Machine Learning

Our rendition of MutantX-S in python. A malware classifier employed using a prototype based clustering approach.

Production-style ML pipeline for click-through rate prediction on the Avazu dataset. 32M training rows, 13M test predictions. L2-regularized logistic regression with smoothed CTR encoding, frequency encoding, and 2^22-dim feature hashing. Validation log loss 0.382 (~11.4% improvement over baseline).