Software
This page collects software developed for my recent papers.
Structured Canonical Correlation Analysis
eccar3 (R package)
The eccar3 package implements canonical correlation analysis through a reduced-rank regression framework. It is designed for high-dimensional settings and supports structured regularization, model selection, and cross-validation.
The package includes cca_rrr() for settings in which one data view is high-dimensional and the other is low-dimensional, as well as ecca() for settings in which both data views are high-dimensional. Available penalties include standard \(\ell_1\) regularization, group-lasso regularization, and total-variation regularization for variables with a known graph structure.
RCCA (R package)
The RCCA package implements regularized canonical correlation analysis with structured penalties. It includes three regularization approaches: standard \(\ell_2\) regularization, partial \(\ell_2\) regularization, and a group penalty for controlling sparsity in high-dimensional settings.
Chromatin 3D Reconstruction
DBMS (R package)
The DBMS package implements distribution-based metric scaling for 3D chromatin reconstruction. It generalizes PoisMS by supporting several probabilistic models for chromatin contact matrices, including Poisson, zero-inflated Poisson, hurdle Poisson, and negative-binomial models.
The package also supports smooth curve estimation through basis control, such as B-splines, and roughness penalization, such as smoothing splines.
PoisMS (R package)
The PoisMS package implements statistical curve-based methods for 3D chromatin reconstruction from contact matrices. It includes three approaches: principal curve metric scaling (PCMS), inspired by classical multidimensional scaling; weighted PCMS (WPCMS), which controls the influence of individual matrix entries; and Poisson metric scaling (PoisMS), which models contact counts using a Poisson distribution.
Low-Rank Matrix Approximation
WLRMA (R package)
The WLRMA package implements weighted low-rank matrix approximation for structured data. It solves rank-constrained optimization problems and their nuclear-norm relaxations using proximal gradient descent with Nesterov and Anderson acceleration. The package also includes a weighted alternating least-squares implementation for high-dimensional matrices.