Boltzmann MapReduce: A Partition-Function Reduce for Forkable Sandboxes

Boltzmann MapReduce: A Partition-Function Reduce for Forkable Sandboxes

Boltzmann MapReduce:用于可分叉沙盒的配分函数归约

Abstract: To leading order under local asymptotic normality (LAN), the confidence density a worker emits over a chunk of size $n$ is a Gibbs—Boltzmann measure $\exp{-\beta E(\theta)}$ whose inverse temperature is the sample size, $\beta=n$.

摘要: 在局部渐近正态性(LAN)下的主导阶中,工作节点针对大小为 $n$ 的数据块所发出的置信密度是一个吉布斯-玻尔兹曼测度 $\exp{-\beta E(\theta)}$,其逆温度即为样本大小,即 $\beta=n$。

Three consequences are exact in the Gaussian/linear case and first-order otherwise: disjoint chunks carry independent Boltzmann factors, so the MapReduce \emph{reduce}, read literally, is a partition function $Z=\int\prod_k h_k,d\theta$ whose mode is precision-weighted (inverse-variance) pooling; frequentist consistency is the zero-temperature limit $T=1/n\to0$.

在高斯/线性情况下,以下三个结论是精确的,在其他情况下则为一阶近似:不相交的数据块携带独立的玻尔兹曼因子,因此从字面上理解,MapReduce 的“归约”(reduce)操作实际上是一个配分函数 $Z=\int\prod_k h_k,d\theta$,其众数是精度加权(逆方差)池化;频率派的一致性即为零温度极限 $T=1/n\to0$。


Paper Details:

  • Authors: Yossi Eliaz
  • arXiv ID: 2607.09689
  • Subjects: Artificial Intelligence (cs.AI); Probability (math.PR); Statistics Theory (math.ST)
  • Submitted: 17 Jun 2026

论文详情:

  • 作者: Yossi Eliaz
  • arXiv ID: 2607.09689
  • 学科分类: 人工智能 (cs.AI);概率论 (math.PR);统计理论 (math.ST)
  • 提交日期: 2026年6月17日