1. The NEWS code for node embeddings in sparse networks
  2. The DIRECT code for classification from few positive examples
  3. The AlphaRob code for constructing portfolios with high out-of-sample Sharpe ratios
  4. The ROLIN code for robust linear classification from limited training data
  5. Community detection in large graphs
    • SVM-Cone (overlapping communities; all low-rank models)
    • SPACL (for the Mixed Membership Stochastic Blockmodel)
    • GeoNMF (non-negative matrix factorization for the Mixed Membership Stochastic Blockmodel)
    • CrossAssociations/AutoPart (parameter-free)
  6. The NetMine package for extracting patterns from large graphs
  7. The F4 non-linear time series forecasting package