Machine learning

Being able to build computer systems that can learn in some sense is one of the very central problems in artificial intelligence. Inspired by certain theoretical advances in the field such as Valiant's PAC model and Rissanen's MDL principle, the project generally aims, unlike traditionally in AI, to apply to machine learning the approach of theoretical computer science and algorithmics. This international trend is called computational learning theory. Our theoretical work is supported by experimenting with the algorithms.

Our results include generalizations and analyses of the RANK algorithm by Ehrenfeucht and Haussler, several theoretical results on the power of the so-called useful and reliable learnability (a variant of the PAC model), and introducing a new family of learning algorithms satisfying the Occam condition for learning decision lists and certain decision trees. The Occam algorithms are suitable for finding common patterns in sets of strings. Two extensive tools for experimenting with learning algorithms have been produced: The TELA system for the study of learning algorithms that use traditional form of the data represented as attribute vectors, and the DALI system for the study of learning algorithms that use strings as the basic data.

The current research themes of the group are:

• Development of a decision tree learning algorithm that has a sound theoretical basis and performs well in practice.

• Development of new MDL learning algorithms for various practical learning tasks as well as development of the complexity theory for MDL learning. Here some very promising results have been obtained recently including the first MDL learning algorithm that has a proven performance guarantee and its application to a clustering problem of biological sequences.

• Further development of the Occam algorithms mentioned above and, in particular, their application to problems in computational biology.

• Further development of the TELA and DALI systems.

The group has good international reputation and is together with nine other European groups a member of the ESPRIT Working Group``NeuroCOLT'' and a site of European Machine Learning Network. The group also works in close co-operation with Prof. Heikki Mannila's data mining group.

The members of the group are Prof. Esko Ukkonen (group leader), Dr. Jyrki Kivinen, Dr. Tapio Elomaa, Timo Lamminjoki, M.Sc Janne Ravantti, Kimmo Valtonen, and M.Sc. Jaak Vilo. The group gets funding from the Academy of Finland and from the European ESPRIT Programme.

Publications: [147, 148, 195-205, 208-215, 221].

Home page: http://www.cs.helsinki.fi/research/pmdm/ml/