Reproducibility in machine learning for health research: Still a ways to go

Matthew B.A. McDermott; Shirly Wang; Nikki Marinsek; Rajesh Ranganath; Luca Foschini; Marzyeh Ghassemi

doi:10.1126/scitranslmed.abb1655

Reproducibility in machine learning for health research: Still a ways to go

Matthew B.A. McDermott, Shirly Wang, Nikki Marinsek, Rajesh Ranganath, Luca Foschini, Marzyeh Ghassemi

Computer Science

Research output: Contribution to journal › Review article › peer-review

Abstract

Machine learning for health must be reproducible to ensure reliable clinical use. We evaluated 511 scientific papers across several machine learning subfields and found that machine learning for health compared poorly to other areas regarding reproducibility metrics, such as dataset and code accessibility. We propose recommendations to address this problem.

Original language	English (US)
Article number	eabb1655
Journal	Science Translational Medicine
Volume	13
Issue number	586
DOIs	https://doi.org/10.1126/scitranslmed.abb1655
State	Published - Mar 24 2021

ASJC Scopus subject areas

General Medicine

Access to Document

10.1126/scitranslmed.abb1655

Cite this

@article{d162dc598f8e4aa7ac5fd4ad01485c72,

title = "Reproducibility in machine learning for health research: Still a ways to go",

abstract = "Machine learning for health must be reproducible to ensure reliable clinical use. We evaluated 511 scientific papers across several machine learning subfields and found that machine learning for health compared poorly to other areas regarding reproducibility metrics, such as dataset and code accessibility. We propose recommendations to address this problem.",

author = "McDermott, {Matthew B.A.} and Shirly Wang and Nikki Marinsek and Rajesh Ranganath and Luca Foschini and Marzyeh Ghassemi",

note = "Publisher Copyright: Copyright {\textcopyright} 2021 The Authors, some rights reserved.",

year = "2021",

month = mar,

day = "24",

doi = "10.1126/scitranslmed.abb1655",

language = "English (US)",

volume = "13",

journal = "Science Translational Medicine",

issn = "1946-6234",

publisher = "American Association for the Advancement of Science",

number = "586",

}

TY - JOUR

T1 - Reproducibility in machine learning for health research

T2 - Still a ways to go

AU - McDermott, Matthew B.A.

AU - Wang, Shirly

AU - Marinsek, Nikki

AU - Ranganath, Rajesh

AU - Foschini, Luca

AU - Ghassemi, Marzyeh

PY - 2021/3/24

Y1 - 2021/3/24

N2 - Machine learning for health must be reproducible to ensure reliable clinical use. We evaluated 511 scientific papers across several machine learning subfields and found that machine learning for health compared poorly to other areas regarding reproducibility metrics, such as dataset and code accessibility. We propose recommendations to address this problem.

AB - Machine learning for health must be reproducible to ensure reliable clinical use. We evaluated 511 scientific papers across several machine learning subfields and found that machine learning for health compared poorly to other areas regarding reproducibility metrics, such as dataset and code accessibility. We propose recommendations to address this problem.

UR - http://www.scopus.com/inward/record.url?scp=85103504047&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85103504047&partnerID=8YFLogxK

U2 - 10.1126/scitranslmed.abb1655

DO - 10.1126/scitranslmed.abb1655

M3 - Review article

C2 - 33762434

AN - SCOPUS:85103504047

SN - 1946-6234

VL - 13

JO - Science Translational Medicine

JF - Science Translational Medicine

IS - 586

M1 - eabb1655

ER -

Reproducibility in machine learning for health research: Still a ways to go

Abstract

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this