Adversarial Learning with Local Coordinate Coding

TL;DR

This paper introduces a Local Coordinate Coding (LCC) sampling method for GANs that leverages learned latent distributions to better preserve data semantics, supported by theoretical bounds and extensive experiments.

Contribution

It proposes a novel LCC-based sampling approach for GANs, improving data semantic preservation and generalization over traditional prior-based methods.

Findings

LCC-based GANs outperform traditional methods on multiple datasets.

A theoretical generalization bound for LCC-GANs is established.

Small input dimensions suffice for effective generalization.

Abstract

Generative adversarial networks (GANs) aim to generate realistic data from some prior distribution (e.g., Gaussian noises). However, such prior distribution is often independent of real data and thus may lose semantic information (e.g., geometric structure or content in images) of data. In practice, the semantic information might be represented by some latent distribution learned from data, which, however, is hard to be used for sampling in GANs. In this paper, rather than sampling from the pre-defined prior distribution, we propose a Local Coordinate Coding (LCC) based sampling method to improve GANs. We derive a generalization bound for LCC based GANs and prove that a small dimensional input is sufficient to achieve good generalization. Extensive experiments on various real-world datasets demonstrate the effectiveness of the proposed method.