On the Structure, Covering, and Learning of Poisson Multinomial Distributions

TL;DR

This paper provides a structural characterization of Poisson Multinomial Distributions (PMDs), showing they are close to a sum of a discretized Gaussian and a smaller PMD, leading to improved covers and learning algorithms.

Contribution

It extends the multidimensional CLT for PMDs, constructs efficient covers, and develops near-optimal learning algorithms for arbitrary dimensions.

Findings

Structural characterization of PMDs as Gaussian plus small PMD

Construction of significantly smaller total variation covers for PMDs

Near-optimal sample complexity for learning PMDs in high dimensions

Abstract

An $(n,k)$-Poisson Multinomial Distribution (PMD) is the distribution of the sum of $n$ independent random vectors supported on the set ${\cal B}_k=\{e_1,\ldots,e_k\}$ of standard basis vectors in $\mathbb{R}^k$. We prove a structural characterization of these distributions, showing that, for all $\varepsilon >0$, any $(n, k)$-Poisson multinomial random vector is $\varepsilon$-close, in total variation distance, to the sum of a discretized multidimensional Gaussian and an independent $(\text{poly}(k/\varepsilon), k)$-Poisson multinomial random vector. Our structural characterization extends the multi-dimensional CLT of Valiant and Valiant, by simultaneously applying to all approximation requirements $\varepsilon$. In particular, it overcomes factors depending on $\log n$ and, importantly, the minimum eigenvalue of the PMD's covariance matrix from the distance to a multidimensional Gaussian random variable. We use our structural characterization to obtain an $\varepsilon$-cover, in total variation distance, of the set of all $(n, k)$-PMDs, significantly improving the cover size of Daskalakis and Papadimitriou, and obtaining the same qualitative dependence of the cover size on $n$ and $\varepsilon$ as the $k=2$ cover of Daskalakis and Papadimitriou. We further exploit this structure to show that $(n,k)$-PMDs can be learned to within $\varepsilon$ in total variation distance from $\tilde{O}_k(1/\varepsilon^2)$ samples, which is near-optimal in terms of dependence on $\varepsilon$ and independent of $n$. In particular, our result generalizes the single-dimensional result of Daskalakis, Diakonikolas, and Servedio for Poisson Binomials to arbitrary dimension.