Funded by the European Commission under contract INCO 977102

Researchers

Liviu Badea, Doina Tilivea, Monica Stanciu, Monica Brinzaru

ILP areas of interest

applications of ILP (for example in knowledge discovery in databases and molecular biology)
refinement operators
knowledge aquisition
representation and inference
concept formation, conceptual clustering
learning recursive clauses from small example sets
optimizing the learned clause set for a given predicate
ILP with constraints

Description of Research

The AI research lab has been involved in several projects in the field of Machine Learning and Knowledge Aquisition.

More specifically, we have developed the ESYS knowledge aquisition environment [Tilivea and Stancioiu] originally based on the personal construct theory, but which has evolved lately into an aquisition and machine learning environment integrating several approaches, such as concept formation, bottom-up as well as top-down conceptual clustering, induction of decision trees and rules from repertory grids and/or databases. Potential applications include fault diagnosis in the chemical industry.

Recently, we have started a project aiming at applying ILP in knowledge engineering. Issues under investigation include

learning recursive clauses from small example sets
optimizing the learned clause set for a given predicate
ILP with constraints
refinement operators

We have also been involved in the field of (logic-based) knowledge representation and reasoning (KR), especially in the field of description/terminological logics (DL) \cite{ECAI96}, where we plan to extend the existing DL systems with learning capabilities.

Our experience with machine learning and knowledge representation has lead us to investigate a unifying approach of KR and ML based on ILP, that would solve the representation problems faced by most of the traditional ML techniques and, on the other hand, would extend the reasoning capabilities of existing KR systems.

We would also like to concentrate on applications of ILP in knowledge discovery in molecular biology databases.

Related Publications

Liviu Badea, Monica Stanciu - Refinement Operators can be (Weakly) Perfect, in Saso Dzeroski and Peter Flach (eds) - Proceedings of the 9th International Conference on Inductive Logic Programming (ILP-99), Bled, 1999.

D. Tilivea, C. Stancioiu ESYS: Interactive Knowledge Acquisition from Multiple Experts Using Personal Construct Theory. Studies in Informatics and Control, Vol. 2, No. 2, June 1993.

D. Tilivea, C. Stancioiu Interactive Knowledge Acquisition from Multiple Experts Using Inductive Learning Techniques (in romanian). Romanian Journal of Informatics and Control, Vol. 5, No. 4, 1995.

Liviu Badea, Monica Stanciu On complete refinement operators, forthcoming.

Liviu Badea, A. Hotaran, M. Brinzaru, M.C. Chidesa Evaluating knowledge engineering methods in Inductive Logic Programming ICI technical report B7-1-96 (in romanian)

Liviu Badea, M. Brinzaru, M.C. Chidesa, C. Toma Testing the capabilities of ILP learning systems ICI technical report A38-2-97 (in romanian)

Mission and objectives of ILPnet2

ILPnet2 is a Network of Excellence consisting of over 20 universities and research institutes. In addition the network actively pursues industrial relations through its End-user-club, consisting of companies and other non-academic institutions interested in practical applications of ILP.

The mission of ILPnet2 is defined by the following long-term objectives:

To co-ordinate ILP research among the nodes of the network.
To promote the co-operation and exchange of research results among the network nodes.
To disseminate information on ILP research and applications to the outside world, including both academic and industrial/non-academic institutions.
To facilitate the transfer of ILP research results to practice.
To support the establishment of the infrastructure necessary for achieving the above objectives.

Research Institute for Informatics (ICI)