Fast nonconvex algorithm for covariate-adjusted precision matrix estimation/conditional Gaussian graphical model estimation

This repository contains our Matlab implementation of covariate-adjusted precision matrix estimation in the paper Covariate Adjusted Precision Matrix Estimation via Nonconvex Optimization.

Useful parameters:

• params.T : Number of total iterations
• params.n : Number of samples
• params.m : Number of output dimensions
• params.d : Number of input dimensions
• params.eta_Gamma : Learning rate for Gamma
• params.eta_Omega : Learning rate for Omega
• params.s_Gamma : Hard-thresholding parameter for Gamma
• params.s_Omega : Hard-thresholding parameter for Omega
• params.Gamma_star : Ground truth matrix Gamma_star, if exists
• params.Omega_star : Ground truth matrix Omega_star, if exists
• params.stopprecision: Threshold for stopping criterion
• params.test : Evaluate test samples (1) or training only (0)
• params.lambda_Gamma : Soft-thresholding parameter for initialize Gamma
• params.lambda_Omega : Soft-thresholding parameter for initialize Omega
• params.epsilon : Ridge parameter for initialize Gamma
• params.nu : Ridge parameter for initialize Omega