Machine Learning Journal: Information for Authors
Contents
Machine Learning publishes papers on a wide range of topics concerning
computational approaches to learning, as indicated in the statement of
Aims and Scope. Research on some of
these topics--specifically, the development and experimental
comparison of learning algorithms and the development and theoretical
analysis of mathematical models of machine learning--has matured to
the point that the Editorial Board has set forth the following
methodological guidelines and recommendations for papers submitted to
Machine Learning on these particular topics.
General Guidelines
- The main exposition of the paper should be aimed at readers who are
generally familiar with machine learning concepts and ideas but not
necessarily with any particular subarea of the field. In particular,
the overall significance of the research results should be
understandable to the general reader.
Guidelines for Experimental Papers
- Papers that introduce a new learning "setting" or type of
application should justify the relevance and importance of this
setting, for example, based on its utility in applications, its
appropriateness as a model of human or animal learning, or its
importance in addressing fundamental questions in machine learning.
- Papers describing a new algorithm should be clear, precise, and
written in a way that allows the reader to compare the algorithm to
other algorithms. For example, most learning algorithms can be viewed
as optimizing (at least approximately) some measure of performance. A
good way to describe a new algorithm is to make this performance
measure explicit. Another useful way of describing an algorithm is to
define the space of hypotheses that it searches when optimizing the
performance measure.
- Papers introducing a new algorithm should conduct experiments
comparing it to state-of-the-art algorithms for the same or similar
problems. Where possible, performance should also be compared against
an absolute standard of ideal performance. Performance should also be
compared against a naive standard (e.g., random guessing, guessing the
most common class, etc.) as well. Unusual performance criteria should
be carefully defined and justified.
- All experiments must include measures of uncertainty of the
conclusions. These typically take the form of confidence intervals,
statistical tests, or estimates of standard error. Proper
experimental methodology should be employed. For example, if "test
sets" are used to measure generalization performance, no
information from the test set should be available to the learning
process.
- Descriptions of the software and data sufficient to replicate the
experiments must be included in the paper. Once the paper has
appeared in Machine Learning, authors are strongly urged to make the
data used in experiments available to other scientists wishing to
replicate the experiments. An excellent way to achieve this is to
deposit the data sets at the Irvine
Repository of Machine Learning Databases. Another good option is
to add your data sets to the DELVE benchmark
collection at the University of Toronto. For proprietary data sets,
authors are encouraged to develop synthetic data sets having the same
statistical properties. These synthetic data sets can then be made
freely available.
- Conclusions drawn from a series of experimental runs should be
clearly stated. Graphical display of experimental data can be very
effective. Supporting tables of exact numerical results from
experiments should be provided in an appendix.
- Limitations of the algorithm should be described in detail.
Interesting cases where an algorithm fails are important in
clarifying the range of applicability of an algorithm.
Guidelines for Theoretical Papers
- The "moral", or general meaning of technical theorems, should be
explained and discussed. Comparisons with general methods in machine
learning should be made.
- The overall consequences of the main theorems should balance the
technical aspects of the paper. That is, a paper that has 30 pages of
detailed mathematics had better have some deep consequences that are
relevant to machine learning at large.
- The proof ideas, and the intuitions behind the proofs of theorems that
are more than routine, should be explained.
Most of the papers published in Machine Learning are regular
papers that give in-depth treatment to a particular topic. However,
Machine Learning also publishes Technical Notes. A
technical note must be a self-contained, small contribution. Often it
is a critique or response to something previously published in
Machine Learning Other times it is a short note describing a
modification or enhancement to an existing algorithm. Many technical
notes could be published as conference papers instead, although even
there they might not be accepted because their significance is often
limited to a small audience.
On the other hand, many conference papers would not be appropriate
technical notes, because their scope is broader and adequate
(non-conference) treatment of the topic requires greater discussion of
previous work, fuller description of experiments (so that they can be
replicated), or complete proofs.
Manuscripts submitted to Machine Learning must be unpublished original
research. If related work has been previously published, the
manuscript submitted to Machine Learning must involve significant
revision or extension. Manuscripts submitted to Machine Learning must
not be concurrently under review at any other journal. If the
manuscript relies heavily on other unpublished manuscripts that are
under review elsewhere, copies of these should be enclosed along with
the manuscript so that reviewers can consult them.
This page maintained by Foster Provost
provost@acm.org