Reproduce “Generic Unsupervised Optimization for a Latent Variable Model With Exponential Family Observables” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Hamid Mousavi, Jakob Drefs, Florian Hirschberger, Jörg Lücke

The authors introduce exponential-family maximum component analysis (E-MCA), where active binary causes influence observations through a maximum, not a sum. They derive EM updates with the same functional form across regular, minimal exponential-family distributions. For one House image with assumed Exponential noise, E-MCA reports 23.41 dB PSNR versus BM3D’s 22.61 dB, not evidence of general superiority.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.