next up previous contents
Next: Using External Knowledge in Up: Computational Information Processing Previous: Computational Information Processing

MENES - Methods for Demanding Neurocomputing Applications

The project is one of the five research projects in the national technology programme "Applications of Learning and Intelligent Systems" coordinated and funded by Technology Development Center (TEKES).

The project was started in 1995 for three years (1.3.1995-28.2.1998), with annual budget of over 2 000 kFIM. The project is financed by (TEKES) and participating enterprises.

Summary of the Project

In neural network applications the solutions are not programmed into the system but rather learned by the networks from example data. The main issues in practical applications are then, how to extract the pertinent information from the data in a form that is relevant in the application, and how to incorporate into the application such necessary information that is not contained in the data.

In the MENES research project methods to tackle these problems are developed and applied in practical applications together with the industrial partners. MENES consists of three subprojects that address different aspects of the general goal.

In subproject Using External Knowledge in Neural Network Models methods for incorporating background knowledge into the neural network models are developed. The approach is based on a set of rules for the desired functional behavior of the solution and algorithms for training the network with both the rules and the data. The project is carried out in the Laboratory of Computational Engineering in Helsinki University of Technology.

In subproject Computationally Intelligent Analysis of Large Data Sets methods for data-analysis with self-organizing map are developed, with special emphasis on relating qualitative data analysis and meanings of the data to the quantitative, numerical analysis produced by the SOM. The project is carried out in the Department of Mathematics in University of Jyväskylä.

In subproject Monitoring and Modeling of Complex Processes methods for monitoring, visualization and prediction of complex industrial processes are developed. The approach is based on the process state analysis with the self-organizing map SOM. The project is carried out in the Laboratory of Computer and Information Science in Helsinki University of Technology.

Juha Merimaa