Revisiting CHAMPAGNE: Sparse Bayesian Learning as Reweighted Sparse Coding

This repo contains the code to reproduce results for our paper "Revisiting CHAMPAGNE: Sparse Bayesian Learning as Reweighted Sparse Coding", currently under review for SampTA 2025.

Getting started

• Install the dependencies using pip install . from the project's root.
• Go to ./benchmarks, and run benchopt run --config config_file. Expect each benchmark to take ~8h to run.
  • The compressed sensing scenario uses configs/config_paper_T=1.yml
  • The MEG setting uses configs/config_paper_lowsnr_2.yml
• The code to reproduce our plots is available in ./plots.