Matthijs van Leeuwen is assistant professor and group leader of the Explanatory Data Analysis group at the Leiden Institute of Advanced Computer Science (LIACS), Leiden University, in the Netherlands. His primary research interest is exploratory data mining: how can we enable domain experts to explore and analyze their data, to discover structure and -ultimately- novel knowledge?
Most of his research concerns the discovery of patterns in data, often using information theoretic principles (minimum description length, maximum entropy). Further, he is interested in integrating expert knowledge by involving the human in the loop.
As he likes working on foundational data mining problems driven by real-world data and problems, he collaborates with partners both in science and industry (e.g., GE Aviation, Honda, Leiden University Medical Center, and Sanquin).
Van Leeuwen has been awarded several grants (NWO Rubicon & TOP, FWO Postdoc, ZonMw Memorabel) and won several best paper awards at international conferences. He co-organised a number of international conferences and workshops, including IDA and IDEA, and co-lectured tutorials on Information Theoretic Methods in Data Mining. He is on the editorial board of DAMI, and on the guest editorial board of the ECML PKDD Journal Track. He was guest editor of a TKDD special issue on Interactive Data Exploration and Analytics.
AG 1, AG 2, AG 3, INET, AG 4, AG 5, SWS, RG1, MMCI
Although mining large numbers of patterns from data is generally easy, finding the right patterns is surprisingly hard. And although nowadays many people have data that they would like to analyze, many do not have the data mining expertise required to do so.
In this talk I will present a framework, paraphrased by the slogan "Mine, Interact, Learn, Repeat", that aims to enable domain experts to find patterns that matter to them by learning their interests through user interaction. In particular, we will discuss an instantiation of this 'human-in-the-loop' framework that combines pattern sampling with preference learning to achieve this goal. For this we build upon existing results from pattern mining, machine learning, information retrieval, and SAT solving.