Reproduce “Predictive Self-Supervised Learning Provably Identifies Stochastic Signals under Nuisance” 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
Fabian A. Mikulasch, Friedemann Zenke
The authors study how predictive learning can distinguish stochastic signal from observation-private nuisance. They combine predictive mutual-information maximization, which preserves information needed for prediction, with latent distribution matching, which constrains its representation. Under assumptions, the signal is identifiable up to an affine map. In a Causal3DIdent image-pair benchmark, object-factor R² scores reach 0.92–0.98; this is not evidence of nuisance removal.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.