This code accompanies our work:
@article{ritchie2020supervised,
title={Supervised PCA: A Multiobjective Approach},
author={Ritchie, Alexander and Balzano, Laura and Kessler, Daniel and Sripada, Chandra S and Scott, Clayton},
journal={arXiv preprint arXiv:2011.05309},
year={2020}
}
There are several helper files. The callable files are:
- lspca_sub.m - supervised pca with the least squares loss function; for regression; uses substitution for \beta in place of alternating updates
- lspca_MLE_sub.m - same as above; uses maximum likelihood updates for tuning parameter \lambda
- lrpca.m - supervised pca with the logistic loss function; for classification
- lrpca_MLE.m same as above; uses maximum likelihood updates for tuning parameter \lambda
- the four files above prepended with 'k' - kernelized versions of those algorithms