EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

3754. Neural style transfer inspired optimisation

Invited abstract in session TD-25: Applications of Machine Learning in Optimization, stream Combinatorial Optimization.

Tuesday, 14:30-16:00
Room: 011 (building: 208)

Authors (first author is the speaker)

1. Stephan Nel
Department of Industrial Engineering, Stellenbosch University

Abstract

In the case of traditional neural style transfer, the aim is to transfer the style embedded within an image, song, or literary work to that of another, e.g. to transfer the art style of Van Gogh’s Starry Night to another image. Contemporary approaches have demonstrated notable efficacy in respect of capturing and translating embedded properties within complex data representations. In this study, an investigation is carried out into the application of neural style transfer to optimisation algorithms. More specifically, the extent to which neural style transfer can capture and transfer salient features of high-quality solutions to other solutions is investigated. The proposed approach entails various modifications to different facets of conventional neural style transfer, such as data set construction, solution encoding, neural architecture, and loss functions. Different benchmark optimisation problems are considered in respect of evaluating the utility of the proposed approach.

Keywords

Status: accepted

Back to the list of papers