deep-significance: Easy and Meaningful Signifcance Testing in the Age of Neural Networks

Dennis Ulmer, Christian Hardmeier, Jes Frellsen

Research output: Contribution to conferencePaperResearchpeer-review

27 Downloads (Pure)

Abstract

A lot of Machine Learning (ML) and Deep Learning (DL) research is of anempirical nature. Nevertheless, statistical significance testing (SST) is still notwidely used. This endangers true progress, as seeming improvements over abaseline might be statistical flukes, leading follow-up research astray while wastinghuman and computational resources. Here, we provide an easy-to-use packagecontaining different significance tests and utility functions specifically tailoredtowards research needs and usability
Original languageEnglish
Publication date2022
Number of pages20
Publication statusPublished - 2022
EventML Evaluation Standards Workshop at the Tenth International Conference on Learning Representations - Virtual Event
Duration: 25 Apr 202229 Apr 2022

Workshop

WorkshopML Evaluation Standards Workshop at the Tenth International Conference on Learning Representations
LocationVirtual Event
Period25/04/202229/04/2022

Fingerprint

Dive into the research topics of 'deep-significance: Easy and Meaningful Signifcance Testing in the Age of Neural Networks'. Together they form a unique fingerprint.

Cite this