Reproduce “Denoising without access to clean data using a partitioned autoencoder” 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

Dan Stowell, Richard E. Turner

The partitioned autoencoder separates signal and noise in field recordings without clean training targets. It reconstructs every example while penalizing designated latent units on noise-only examples, then reconstructs the signal from the other units. In a birdsong test, it outperformed a standard denoising autoencoder on truly clean targets when noise sources matched, but not when they were unmatched.

Summary by alphaXiv.

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