JMLR Information for Authors


The Journal of Machine Learning Research will publish original, unpublished research in Machine Learning reporting In addition, extremely well-written surveys of existing work will also be considered for publication.


Papers must be concise and complete; manuscripts should be carefully proofread and polished. Submissions that do not meet these criteria may be returned unreviewed.

All claims should be clearly articulated and supported either by empirical experiments or theoretical analyses. When appropriate, authors are encouraged to implement their work and to demonstrate its utility on significant problems; any experiments reported should be reproducible. Papers describing systems should clearly describe the contributions or the principles underlying the system. Papers describing theoretical results should also discuss their practical utility. In general, it should be clear how the work advances the current state of understanding and why the advance matters. Papers should report on what was learned in doing the work, rather than merely on what was done.

Authors must clearly acknowledge the contributions of their predecessors. If a paper introduces new terminology or techniques, it should also explain why current terminology or techniques are insufficient.


Submitted work cannot have been published previously or be pending publication in another journal. We will consider research that has been published, or is under consideration for publication at workshops or conferences. In these cases, we expect the JMLR submission to go into greater depth and extend the published results in a substantive way. Authors must notify JMLR about previous or pending conference publication at the time of submission.

Submission procedure

JMLR accepts submissions via its own electronic submission management system.

Articles must be submitted in electronic form; either PostScript or PDF is acceptable. Submissions should be in single column format, typeset in 11 point font or larger, and should include all author contact information on the first page. Although not required, we encourage authors to use the JMLR format (described in the authors guide).

Articles may be accompanied by online appendices containing data, demonstrations, instructions for obtaining source code, or the source code itself  if appropriate. We strongly encourage authors to include such appendices along with their papers. (Note: if an online appendix contains source code, we will require you to sign a release form prior to publication freeing us from liability.)

To submit a paper, please

  1. Prepare it in PDF, Postscript, or Microsoft Word format (if the submission contains multiple files, please create an archive in tar or zip format).
  2. Ensure that the file to be uploaded is less than 5Mb in size.
  3. Rename the file so that its extension indicates the type (.ps or .pdf).
  4. Ensure that the title page contains
  5. Go to, register and log in.

You can then confirm that your manuscript has been received by selecting (at the top of the page) "manuscript listing" followed by "author." Your manuscript's ID tag should appear along the left-hand side. Following its link will give the current status of the manuscript.


JMLR has a commitment to rigorous yet rapid reviewing. In special circumstances (submission is unusually long or complex, or is received during the review period of a major conference), a longer review period may be necessary. In this case, authors will be notified of the required review period as soon as practical.

JMLR reviewers are instructed to make recommendations to accept or reject. If a paper requires substantial revisions, it will not be accepted. In this case, the editor may recommend that the author submit a revised version, but there is no guarantee it will be reviewed by the same referees, or that it will eventually be published. A paper may be revised and resubmitted at most once. To avoid misunderstandings about interpretation of reviewer and editor recommendations, we encourage authors to contact the editor prior to resubmission with any salient questions.

If a paper is accepted, the editor and reviewers may require minor revisions. The author should return the revised paper as soon as possible. Normally, authors will have up to three months to accomplish their revisions. Upon receiving a revised paper the editor may, at their discretion, ask the reviewers to re-review the revised paper. Electronic publication will occur immediately upon receipt of the final version of an accepted article.


When a paper is accepted, authors retain ownership of the copyright, but must sign a copyright agreement form giving MIT Press the right to publish the paper. Authors are encouraged to provide HTML versions of the paper, as well as electronic forms of any applicable data sets or executable code. Authors are encouraged to add links to the article and the JMLR site to their personal web pages; if the authors wish to create a page pointing to subsequent related research, JMLR will maintain a link from the original paper to this page. See the authors guide for a detailed checklist of post-acceptance actions.

A quarterly paper volume will be published and sold to libraries and individuals.

Papers that fall in the area of support vector machines and kernel methods will be considered for joint publication in JMLR and on If appropriate, please indicate in your cover letter that you would like your submission to be considered for inclusion in the kernel section. Further information can be found in the JMLR section of the kernel machines web repository.

After final acceptance, authors are encouraged to create an HTML version of their article to accompany the "official" PostScript (or PDF) article. The HTML version should be identical in content to the PostScript (or PDF) article. Production of HTML versions may be aided by automatic tools available, such as the LaTex2HTML translator. Authors also may publish "forward pointers" to subsequent articles on the same topic by contacting the executive editor.

Special Topics

JMLR is interested in publishing special issues, because it is often useful to have a collection of strong papers on a single topic collected together in an accessible way, and because the deadlines of special issues are often a useful forcing function for people to finally write up their journal papers.

Regular special issues, in which papers appear contiguously in the printed version, have proved to be a major logistical nightmare, largely because we have to wait until the last final version is submitted before we can publish any of them. There are also some packing problems associated with trying to get all the papers of a special issue into a single physical volume.

For these reasons, we have decided, in future, to have special topics instead of special issues. These will operate in essentially the same mode as special issues, except that they papers will not appear contiguously in print. They will be be published individually as soon as the first one is ready, and will be incrementally collected into a special page on the web site.

Proposing a Special Topic

If you'd like to propose a special topic, please submit a short proposal outlining: The action editors will discuss and make accept/reject decisions on proposals.


Here are some things we have observed about special issues, which are useful to keep in mind.

Special issues tend to have somewhat lower quality papers than regular issues. This may be because: it's a somewhat more closed community that is reviewing and making decisions; or that papers are often invited by the guest editors, who then find it difficult to reject marginal ones. In any case, it's important to be aware of this phenomenon and try to keep it in check.

Special issues almost always run late. Sometimes very drastically so. Guest editors almost always underestimate the amount of work required to manage the issue and the difficulty of extracting reviews from even the best-intentioned of their colleagues.

Links - authors guide, forms and style files

Links and portions of this text reproduced by kind permission of the Journal of Artificial Intelligence Research.

Copyright © JMLR 2000. All rights reserved.