arXiv:2512.13997v2 Announce Type: replace-cross Abstract: Existing two-sample testing techniques, particularly those based on choosing a kernel for the Maximum Mean Discrepancy (MMD), often assume equal sample sizes from the two distributions. Applying these methods in practice can require discarding valuable data, unnecessarily reducing test power. We address this long-standing limitation by extending the theory of generalized U-statistics and applying it to the usual MMD estimator, resulting in new characterization of the asymptotic distributions of the MMD estimator with unequal sample size
The continuous evolution of AI and machine learning techniques demands more robust statistical methods for foundational tasks like two-sample testing, especially as datasets become more diverse in size and structure.
Improved statistical testing methods for unequal sample sizes enhance the reliability and efficiency of AI model validation, potentially reducing data waste and increasing statistical power in various applications.
The ability to accurately compare two distributions with unequal sample sizes without discarding data becomes more practical, offering more robust and efficient statistical inference in machine learning and data science.
- · AI/ML researchers
- · Data scientists
- · Clinical trials
- · A/B testing platforms
- · Inefficient statistical methods
- · Studies constrained by equal sample size assumptions
More accurate and powerful statistical comparisons become feasible in machine learning research and applications.
This could lead to more efficient use of experimental data, particularly in fields where data collection is expensive or limited.
Consequential improvements in the reliability of scientific discoveries and the performance of data-driven systems leveraging these enhanced testing capabilities.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG