Discussions
Discussions were organised in groups, on 5 different topics. The topics are described below, the groups can be found here.
A. Pattern selection
This discussion is dedicated to the general question of mining or selecting the right patterns, in order to prevent reporting a flood of patterns or sets of patterns that overfit the data. Typical subjects are Pattern Set Mining, Pattern (Set) Validation and the general idea of Useful Patterns.
B. Subgroup Discovery meets Pattern Mining
This topic concerns the differences and potential overlap between the fields of Subgroup Discovery and Pattern Mining. The aim is to identify commonalities, overcome irrelevant differences in terminology, and provide a vision for the unison of the two fields, including future developments.
C. Scalability
This discussion considers scalability issues that stem from complex modelling tasks, as well as large and complex datasets. Especially when dealing with data beyond the basic itemset data, smart algorithms will need to be developed to allow practical application. Additionally, methods for heuristic and non-exhaustive discovery methods will be considered.
D. Applications
This topic concerns the various applications of SD en PM, such as Bioinformatics, Stream Mining, Process Mining and Social Networks. Specifically, applications that form an inspiration for new discovery algorithms are of interest.
E. Mining complex data
This discussion will focus on the discovery in complex data, such as non-tabular, relational or dynamic data. Furthermore, specific attention will be given to data with complex target concepts, such as multi-label data and other tasks that involve multiple target attributes. Methods will be considered that find patterns that model the dependencies between the target attributes.