Policy ForumReproducibility

Certify reproducibility with confidential data

See allHide authors and affiliations

Science  12 Jul 2019:
Vol. 365, Issue 6449, pp. 127-128
DOI: 10.1126/science.aaw2825

eLetters is an online forum for ongoing peer review. Submission of eLetters are open to all. eLetters are not edited, proofread, or indexed.  Please read our Terms of Service before submitting your own eLetter.

Compose eLetter

Plain text

  • Plain text
    No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
Author Information
First or given name, e.g. 'Peter'.
Your last, or family, name, e.g. 'MacMoody'.
Your email address, e.g. higgs-boson@gmail.com
Your role and/or occupation, e.g. 'Orthopedic Surgeon'.
Your organization or institution (if applicable), e.g. 'Royal Free Hospital'.
Statement of Competing Interests

This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.

Vertical Tabs

  • Solving reproducibility problems is not so easy

    Christophe Pérignon et al. wrote an article entitled “Certify reproducibility with confidential data” (1). Since the advent of GPU (graphic processing unit) technology, a large number of parallel processing cores can run for machine learning. There are useful machine learning frameworks including Chainer, Keras, Pytorch, and others for building AI machine learning systems. However, many of these frameworks cannot guarantee the reproducibility because of pseudorandom numbers spread in multiple GPUs. Pytorch is one and only one machine learning framework to be able to reproduce the same result in multiple GPUs environment as long as complex pseudorandom numbers are properly initialized. It is not a matter of confidential data or not. Therefore, it is extremely difficult to reproduce the same result in the multiple GPUs environment so that we are not sure how Christophe Pérignon et al. can implement certifying reproducibility with confidential data.

    1. Christophe Pérignon et al., Certify reproducibility with confidential data, Science 12 Jul 2019: Vol. 365, Issue 6449, pp. 127-128

    Competing Interests: None declared.

Stay Connected to Science