With and without directionality(Nora)
- SVD with numbers of interactions(Nora) http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.TruncatedSVD.html
try with n_components = ~40,000 save out explained_variance_ratio_ : array, [n_components] and plot it to see best ones from those best one use that number as n_components and run again read out components small matrix: transpose * components = big matrix for each data point do row* column to get value truncate between 0 and 1
- Poisson-gamma matrix factorization model(non-negative factorization model)(Nora) http://scikit-learn.org/stable/auto_examples/applications/topics_extraction_with_nmf.html
- Mixture model (Nick) http://scikit-learn.org/stable/modules/classes.html#module-sklearn.mixture
Outputs: roc file(csv): fpr. tpr probs(csv): test value, probability
Create pretty graphs(Nick)
Results: violin plots ROC curves
Nick's Current Agenda Items: -Perform 2D plot of data clusters -Derive # of cluster centers from plot -Run GMM -Derive probabilities -Plot ROC curves, etc.