# Rethinking Generative Mode Coverage: A Pointwise Guaranteed Approach

arXiv: https://arxiv.org/abs/1902.04697 · HTML: https://mlsanity.com/paper/1902.04697/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 13 Feb 2019.

## Built on by (most cited first, 3 of 3)

- [Leverage Score Sampling for Complete Mode Coverage in Generative Adversarial Networks](https://mlsanity.com/paper/2104.02373/) (2021)
- [VARGAN: Variance Enforcing Network Enhanced GAN](https://mlsanity.com/paper/2109.02117/) (2021)
- [DualVAE: Controlling Colours of Generated and Real Images](https://mlsanity.com/paper/2305.18769/) (2023)

## Abstract (arXiv)

Many generative models have to combat textitmissing modes. The conventional wisdom to this end is by reducing through training a statistical distance (such as f-divergence) between the generated distribution and provided data distribution. But this is more of a heuristic than a guarantee. The statistical distance measures a textitglobal, but not textitlocal, similarity between two distributions. Even if it is small, it does not imply a plausible mode coverage. Rethinking this problem from a game-theoretic perspective, we show that a complete mode coverage is firmly attainable. If a generative model can approximate a data distribution moderately well under a global statistical distance measure, then we will be able to find a mixture of generators that collectively covers textitevery data point and thus textitevery mode, with a lower-bounded generation probability. Constructing the generator mixture has a connection to the multiplicative weights update rule, upon which we propose our algorithm. We prove that our algorithm guarantees complete mode coverage. And our experiments on real and synthetic datasets confirm better mode coverage over recent approaches, ones that also use generator mixtures but rely on global statistical distances.