Objective
The principal goal of the project is to develop and investigate pattern recognition systems based on a high-performance neural network classifier that combines fast learning and recognition, high performance, and ability to form arbitrary complex decision boundaries. A prototype of toolkit for pattern recognition systems development will be created which will enable the development of pattern recognition applications superior to existing systems in respect of high performance and the possibility to optimize feature sets.
The project will benefit from the original architecture of neural networks on which the classifier is based and efficient implementation through parallel transputer systems.
The project will contribute to progress in the area of pattern recognition. The neural network classifier to be developed will substantially assist the solution of traditional classification problems, including financial analysis, automated medical tests, industrial inspection, diagnostics, etc. Moreover, it will enable a number of important problems to be solved that are difficult for other classifiers including pattern classification tasks for large training sets (>100000 training samples), short recognition time (microseconds) and optimization of feature sets
Topic(s)
Call for proposal
Data not availableFunding Scheme
Data not availableCoordinator
09126 Chemnitz
Germany