This workshop follows up on the first workshop success with around 40 participants, 18 submissions, 13 regular papers and 3 work-in-progress contributions. This second edition will focus on reusing models. Indeed when a change of context is observed during deployment, it is often hard to train a new model, while there exist models trained in different conditions. This year a challenge competition on real data (bike sharing) will be organised to promote models reuse over multiple contexts and foster comparisons and discussions.
LMCE 2015 workshop is organised in conjunction with MoReBikeS Discovery Challenge (MoReBikeS in the morning and LMCE in the afternoon). Schedule for both events can be found here.
Reuse of learnt knowledge is of critical importance in the majority of knowledge-intensive application areas, particularly because the operating context can be expected to vary from training to deployment. In machine learning this has been most commonly studied in relation to variations in class and cost skew in classification. While one classifier outputting class labels may be sufficient and highly specialised for one particular operating context, it may not behave well for other particular operating contexts. Instead of training several specialised models for each particular operating context, it is usually more cost-effective to learn one general, versatile model, for example in the form of a scoring classifier outputting scores or probabilities, that can be successfully applied to several contexts through an appropriate procedure, such as the choice of a threshold.
This successful approach can be generalised to many other problems and areas in machine learning, where models are meant to be more general and more parameterisable, so they can stand up better for changes in the data distribution, data representation, associated costs, noise, reliability, background knowledge, etc. With this view, models are not continuously retrained and assessed every time a change happens, but kept, enriched and validated in a longer life cycle. This idea is known as 'reframing', defined as the process of adapting a model to perform well on a range of operating contexts. This is usually performed by the use of a more versatile model, which is not discarded after each application.
This workshop is intended to improve our understanding of the principles of model generalisation and reuse, and practical ways of integrating the two. Particular emphasis will be given to foster the cross-fertilisation of ideas and applications with related areas in machine learning. Consequently, the presentations and discussions will be open to a broad list of topics (not exhaustive):
The workshop length is half a day divided into 3 sessions. LMCE will begin with a short presentation and an invited talk by Carlos Soares tittled "Metalearning for model management: beyond model reuse".
Each session will consist of short paper presentations, devoting an important share of time to informal discussion and interaction between the participants. The workshop will be closed with an open discussion about more promising open problems and research areas of the workshop topics, continuation of the workshop, future related events, etc.
Attendance could be limited to 40 people, with preference given to people who are presenting papers.
Organisation will provide a computer with dedicated software (Powerpoint and Acrobat Reader) and a video projector for the computer.
We welcome submissions describing work in progress as well as more mature work related to learning over multiple contexts. Submissions should be between 6 and 16 pages in the same format as the main conference (LNAI, single-blind revision).
Papers will be selected by the program committee according to the quality of the submission and its relevance to the workshop topics. All accepted papers will be published on the workshop web site. The publication of a selected set of papers for a special volume or a journal issue is considered, but this will depend on the success and overall results of the workshop.
Submission website: https://www.easychair.org/conferences/?conf=lmce2015
LMCE 2015 ACCEPTED PAPERS can be found here.
