Copula & Marginal Flows: Disentangling the Marginal from its Joint

TL;DR

This paper introduces copula and marginal flows (CM flows) that enable exact modeling of distribution tails in deep generative networks, addressing a key limitation in understanding their distributional properties.

Contribution

It derives upper bounds for generative network tails and proposes CM flows for precise tail modeling, a novel approach in deep learning.

Findings

CM flows accurately model distribution tails.

Upper bounds demonstrate limitations of optimal generative networks.

Numerical results validate the effectiveness of CM flows.

Abstract

Deep generative networks such as GANs and normalizing flows flourish in the context of high-dimensional tasks such as image generation. However, so far exact modeling or extrapolation of distributional properties such as the tail asymptotics generated by a generative network is not available. In this paper, we address this issue for the first time in the deep learning literature by making two novel contributions. First, we derive upper bounds for the tails that can be expressed by a generative network and demonstrate Lp-space related properties. There we show specifically that in various situations an optimal generative network does not exist. Second, we introduce and propose copula and marginal generative flows (CM flows) which allow for an exact modeling of the tail and any prior assumption on the CDF up to an approximation of the uniform distribution. Our numerical results support the use of CM flows.