statistics

This paper proves the Ho–Nguyen conjecture on the order of density cancellation in over-fitted Gaussian mixtures for all cluster sizes M≥2, deriving exact maximum-likelihood exponents and local minimax rates, and applies the coefficient framework to identifiability issues in mixtures of experts.

本文提出一种针对小样本场景下缺失质量与未见物种计数估计的“脆弱性感知”正则化框架，通过严格凹的熵-KL优化提供闭式解，作为对 Good-Turing 等经典估计器的轻量级修正，在长尾、单例主导的小样本场景中显著提升稳定性。

AI appears to be everywhere, yet 98% of U.S. households have not yet paid for it, reflecting that consumer-side monetization is still in its early stages.

本文研究模型误设下的有界回归问题，分析了 sign-betting regression 估计器在 L1 度量下的高概率 oracle 不等式，并证明了 proper learning 的最优通用近似因子为 3。

Show HN: an interactive web experiment that tests how random users' choices really are, hosted at lab.specr.net's rigged-draw page.

Eli Bendersky's blog post explains Monte Carlo simulations through simple, hands-on examples, including estimating pi by dart-throwing and simulating the card game SET to answer combinatorial probability questions.

A Breakthrough Journal critique argues that a series of Harvard studies linking proximity to nuclear power plants with higher cancer mortality rest on flawed 'proximity score' methodology, demonstrating that the same approach absurdly attributes thousands of cancer deaths to living near Costcos, colleges, or baseball stadiums.

Velocity Radar is a tool that measures true GitHub momentum across public and private repos, with features for pace, breakdown, and stamina, designed for developers and AI agent commanders.

The article argues against the existence of a loneliness epidemic due to statistical misconceptions and highlights scurvy as a more significant issue.

This study benchmarks coding agents on reproducing Eurostat statistics, finding that semantic validation and a retry budget are crucial for reliability, not just execution diagnostics.

Naoki Egami, an associate professor at MIT, researches the integration of AI and statistical methods in political science to enhance the accuracy of social science studies.

The article presents modern frequentist statistical methods through fake-data simulation, providing a practical framework for data analysis.

MIT CSAIL shares a beginner's guide to statistics, providing foundational educational content for data science and AI.

A free 490-page textbook on causal inference by Peng Ding, covering topics from correlation to longitudinal data with accompanying R code and datasets.

This paper describes two pre-specified empirical tests to evaluate whether phase-derived features improve prediction accuracy for commitment or contradiction endpoints, but results show no significant improvement over baseline models.

A live dashboard that counts and displays AI-related headlines on Hacker News, resetting the timer each time a new AI headline appears.

This article discusses statistical methods to predict the release dates of AI models, likely analyzing trends and patterns in development cycles.

An analysis of fix-commit proportions in popular GitHub repositories finds an average fix-ratio of about 10%, slightly lower than the author's previous observation of 15% in production software.

This paper critiques the pessimistic meta-inductive argument against scientific realism by leveraging convergence concepts from frequentist statistics and machine learning, showing that ordinary induction achieves convergence while meta-induction fails.

This tweet announces the second edition of 'Introduction to Modern Statistics,' an open-source textbook available for free online or at a name-your-own-price PDF.