Adversarial Autoencoders in Operator Learning

TL;DR

This paper investigates the integration of adversarial training into neural operator architectures like DeepONets and Koopman autoencoders to enhance their performance in operator learning tasks.

Contribution

It introduces the novel application of adversarial additions to neural operator architectures, demonstrating potential performance improvements.

Findings

Adversarial addition improves autoencoder performance.

Enhanced accuracy in operator learning tasks.

Potential for broader application in neural operators.

Abstract

DeepONets and Koopman autoencoders are two prevalent neural operator architectures. These architectures are autoencoders. An adversarial addition to an autoencoder have improved performance of autoencoders in various areas of machine learning. In this paper, the use an adversarial addition for these two neural operator architectures is studied.