We propose a new model-free ensemble classification framework, RaSE algorithm, for the sparse classification problem. In RaSE algorithm, for each weak learner, some random subspaces are generated and the optimal one is chosen to train the model on the basis of some criterion. To be adapted to the problem, a novel criterion, ratio information criterion (RIC) is put up with based on Kullback-Leibler divergence. Besides minimizing RIC, multiple criteria can be applied, for instance, minimizing extended Bayesian information criterion (eBIC), minimizing training error, minimizing the validation error, minimizing the cross-validation error, minimizing leave-one-out error.

Sure Independence Screening

Given a sample of class 0 and class 1 and a classification method, the package generates the corresponding Neyman-Pearson classifier with a pre-specified type-I error control and Neyman-Pearson Receiver Operating Characteristic (NP-ROC) Bands.
R package. Relevant paper. I’ve also made a Youtube video as follows.

Provides an efficient procedure for fitting the entire solution path for high-dimensional regularized quadratic generalized linear models with interactions effects under the strong or weak heredity constraint.
R package. Relevant paper.

Powered by the Academic theme for Hugo.