graper

graper enables regression and classification on high-dimensional data with adaptively chosen relative strengths of penalization for different feature groups, such as different assays or omic types, using a variational Bayes approach.

Key Features:

• Adaptive group penalization: The optimal relative strengths of penalization for different feature groups are chosen adaptively.
• Support for external covariates: The method incorporates external covariates in the regression and classification framework.
• Variational Bayes optimization: Parameter estimation and optimization are performed using a variational Bayes approach.
• Multi-omic data integration: Different feature groups can represent different assays or omic types.

Scientific Applications:

• High-dimensional regression: Regression analysis on high-dimensional datasets with grouped features.
• High-dimensional classification: Classification tasks on high-dimensional datasets with grouped features.
• Multi-omic data analysis: Integration and analysis of data from different assays or omic types with differential penalization.

Methodology:

Optimization is performed using a variational Bayes approach to estimate adaptive relative penalization strengths across feature groups.