SLIM

SLIM is a machine learning method to build optimized data-driven scoring systems. SLIM uses integer programming to fit models that are fully optimized for accuracy, sparsity, and integer coefficients. The models can handle non-trivial constraints without parameter tuning (e.g., limits on model size, sensitivity, specificity).

slim-python | slim-matlab

RiskSLIM

RiskSLIM is a machine learning method to build optimized data-driven risk scores. RiskSLIM uses integer programming to fit models that are fully optimized for sparsity and integer coefficients, and that handle non-trivial constraints without parameter tuning. RiskSLIM uses a new cutting-plane method so that training can scale to datasets with a huge number of samples.

risk-slim

Classification Pipeline

Train binary classification models using popular methods in R.

classification-pipeline

Group'em

Create student groups with preferences for in-class activities online.

David Bau, Nick Sledeski, Shruti Banda.

demo | group-em

This site was a group project for the UI design class at MIT