Information Theory Considerations in Patch-based Training of Deep Neural Networks on Seismic Time-Series

Jesper Sören Dramsch, Mikael Lüthje

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

137 Downloads (Pure)

Abstract

Recent advances in machine learning relies on convolutional deep neural networks. These are often trained on cropped image patches. Pertaining to non-stationary seismic signals this may introduce low frequency noise and non-generalizability.
Original languageEnglish
Title of host publicationProceedings of the First EAGE/PESGB Workshop on Machine Learning (London2018)
PublisherEuropean Association of Geoscientists and Engineers
Publication date2018
Pages46-48
ISBN (Print)978-1-5108-7668-2
DOIs
Publication statusPublished - 2018
EventFirst EAGE/PESGB Workshop Machine Learning - Olympia London, London, United Kingdom
Duration: 29 Nov 201830 Nov 2018

Conference

ConferenceFirst EAGE/PESGB Workshop Machine Learning
LocationOlympia London
CountryUnited Kingdom
CityLondon
Period29/11/201830/11/2018

Fingerprint Dive into the research topics of 'Information Theory Considerations in Patch-based Training of Deep Neural Networks on Seismic Time-Series'. Together they form a unique fingerprint.

Cite this