Reproduce “Prediction Under Imperfect Compression: A Theory of Approximate MDL” 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

Qian Li, Xinyu Mao, Shang-Hua Teng, Guangxu Yang

The authors study whether approximate minimum description length (MDL) selection supports reliable sequential prediction. For countable model classes containing the true measure, every selector within fixed additive score error has finite cumulative expected squared prediction error when regularization lambda is at least 1. Some selectors fail below 1; multiplicative approximation admits infinite-error counterexamples at every positive regularization strength.

Summary by alphaXiv.

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