Home > Research > Publications & Outputs > Using J-K-fold Cross Validation to Reduce Varia...

Electronic data

  • coling2018

    Accepted author manuscript, 452 KB, PDF document

    Available under license: CC BY: Creative Commons Attribution 4.0 International License

View graph of relations

Using J-K-fold Cross Validation to Reduce Variance When Tuning NLP Models

Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSNConference contribution/Paper

E-pub ahead of print
Publication date06/2018
Host publicationProceedings of COLING 2018
Number of pages12
Original languageEnglish
EventConference on Computational Linguistics - Santa Fe Community Convention Center, Santa Fe, United States
Duration: 20/08/201826/08/2018
Conference number: 27
https://coling2018.org/

Conference

ConferenceConference on Computational Linguistics
Abbreviated titleCOLING
CountryUnited States
CitySanta Fe
Period20/08/1826/08/18
Internet address

Publication series

NameProceedings of COLING 2018
PublisherAssociation for Computational Linguistics
ISSN (Print)1525-2477

Conference

ConferenceConference on Computational Linguistics
Abbreviated titleCOLING
CountryUnited States
CitySanta Fe
Period20/08/1826/08/18
Internet address

Abstract

K-fold cross validation (CV) is a popular method for estimating the true performance of machine learning models, allowing model selection and parameter tuning. However, the very process of CV requires random partitioning of the data and so our performance estimates are in fact stochastic, with variability that can be substantial for natural language processing tasks. We demonstrate that these unstable estimates cannot be relied upon for effective parameter tuning. The resulting tuned parameters are highly sensitive to how our data is partitioned, meaning that we often select sub-optimal parameter choices and have serious reproducibility issues.

Instead, we propose to use the less variable J-K-fold CV, in which J independent K-fold cross validations are used to assess performance. Our main contributions are extending J-K-fold CV from performance estimation to parameter tuning and investigating how to choose J and K. We argue that variability is more important than bias for effective tuning and so advocate lower choices of K than are typically seen in the NLP literature, instead use the saved computation to increase J. To demonstrate the generality of our recommendations we investigate a wide range of case-studies: sentiment classification (both general and target-specific), part-of-speech tagging and document classification.

Bibliographic note

COLING 2018. Code available at: https://github.com/henrymoss/COLING2018