Research map / Learning theory and algorithms / Sample complexity
Entropy Rate Estimation for Markov Chains with Large State Space
Yanjun Han, Jiantao Jiao, Chuan-Zheng Lee, Tsachy Weissman, Yihong Wu, Tiancheng Yu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use a minimax-optimal Shannon entropy estimator within the conditional entropy-rate approach instead of empirical entropy | Entropy-rate estimation on stationary reversible Markov chains · Stated mixing regime | absolute estimation error: C'\frac{S^2}{n} -> C\frac{S^2}{n\ln S} | Improved |
| Use the conditional JVHW entropy estimator to estimate 4-gram conditional entropy | English-language entropy estimation · Google One Billion Words (1BW) | conditional entropy (bits per word): 3.46; current state-of-the-art cross-entropy rate: 4.55 | Mixed |
| Use the conditional JVHW entropy estimator on the PTB corpus | English-language entropy estimation · Penn Treebank (PTB) | conditional entropy (bits per word): 1.50; state-of-the-art cross-entropy rate: 5.96 | Mixed |
| Use empirical conditional entropies rather than a minimax-optimal entropy estimator | Entropy-rate estimation on Markov chains · Large state space, stated mixing regime | sample complexity for consistency: \frac{S^2}{\ln S} -> S^2 | Hurt |
| Estimate the entropy rate without enough samples to cover the transition-state parameters | Empirical entropy rate · Memoryless uniform chain | bias: \ln \left(\frac{S^2}{n+ S-1} \right) | Hurt |
Closest papers by meaning
- Estimating Entropy of Distributions in Constant Space (NeurIPS 2019 · similarity 0.88)
- Entropy testing and its application to testing Bayesian networks (NeurIPS 2024 · similarity 0.83)
- On the Entropy Calibration of Language Models (NeurIPS 2025 · similarity 0.83)
- Efficient Perplexity Bound and Ratio Matching in Discrete Diffusion Language Models (ICLR 2025 · similarity 0.83)
- Chain of Log-Concave Markov Chains (ICLR 2024 · similarity 0.83)
- Conservation Laws for Diffusion Models (NeurIPS 2026 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
