Ordered Homogeneity Pursuit Lasso for Group Variable Selection
Generate simulation data for benchmarking sparse regressions (Gaussian...
Cross-validation for Ordered Homogeneity Pursuit Lasso
Compute D, L, and C in the Fisher optimal partitions algorithm
Fisher optimal partition
OHPL: Ordered Homogeneity Pursuit Lasso for Group Variable Selection
Ordered Homogeneity Pursuit Lasso
Compute RMSEP, MAE, and Q2 for a test set
Make predictions based on the fitted OHPL model
Extract the prototype from each variable group
Ordered homogeneity pursuit lasso (OHPL) algorithm for group variable selection proposed in Lin et al. (2017) <DOI:10.1016/j.chemolab.2017.07.004>. The OHPL method exploits the homogeneity structure in high-dimensional data and enjoys the grouping effect to select groups of important variables automatically. This feature makes it particularly useful for high-dimensional datasets with strongly correlated variables, such as spectroscopic data.
Useful links