Special Session on Interpretable Systems at IEEE SSCI 2013
par Fabrice Rossi
Michael Biehl (Intelligent Systems Group, University of Groningen) and Fabrice Rossi (SAMM, Université Paris 1) organize a special session on Interpretable systems in machine learning, data analysis, and visualization at the CIDM 2013 conference part of the IEEE SSCI 2013 conference in Singapore.
Call for paper (
submission deadline 10 Oct 2012 new (final) deadline 23 Nov 2012) :
The quality of machine learning systems is, most frequently, evaluated in terms of quantitative measures such as the classification error, approximation accuracy or prediction. These criteria are perfectly appropriate where working phase performance is the principal goal of the data analysis.
However, in many application areas the interpretation of the systems in use constitutes an aspect of increasing importance. Plausible, intuitive approaches facilitate useful discussions with the domain experts and can provide important insights into the nature of the data and the problem at hand.
The acceptance of methods developed for real world data analysis, hinges critically on their comprehensibility and transparency for the domain expert. For this reason, simple (e.g. linear) systems with suboptimal performance are often preferred over highly sophisticated (e.g. non-linear) models in practical contexts. Methods with an apparent blackbox character complicate communication between machine learning and domain experts and hinder the dissemination of novel approaches.
This special session addresses (but is not limited) to the following topics :
Plausible and interpretable systems for supervised and unsupervised learning
Interpretation of non-linear systems for classification, regression and unsupervised data analysis
Objective evaluation and comparison of interpretability in machine learning
Visualization of complex and high-dimensional real world data
Feature weighting and selection schemes
Prototype and similarity based systems
Interaction with domain experts in application examples
Generative models for the analysis of real world data
Submissions should be done according the general instructions of the conferences, before the deadline (
10 Oct 2012 23 Nov 2012).