Overcome neural limitations for real world applications by providing confidence values for network predictions


Contact
lkindermann [ at ] awi-bremerhaven.de

Abstract

In this paper we present an incremental construction algorithmfor continuous learning tasks and one of itsspecial features - simultaneous learning of the targetfunction and a confidence value for the system predictions.The basis of the hybrid system is a radial basisfunction (RBF) network layer. The second layer consistsof local models. The two layers are closely combinedwith a strong interaction. The number of RBFneuronsand the number of local models have not to bedetermined in advance. This is one of the main advantagesof the algorithm. Another advantage emphasizedin this paper is the ability to learn the training data distributionsimultaneously to the learning of the targetfunction. The learned data set distribution can be usedas a confidence value for a given network prediction.The development of the described approach is embeddedin a larger project that is primarily concerned withsystem identification tasks for industrial control such assteel processing.



Item Type
Conference (Conference paper)
Authors
Divisions
Programs
Peer revision
Not peer-reviewed
Publication Status
Published
Event Details
Proceedings of the Sixth International Conference on Neural Information Processing (ICONIP'99), Perth.
Eprint ID
10403
Cite as
Tagscherer, M. , Kindermann, L. , Lewandowski, A. and Protzel, P. (1999): Overcome neural limitations for real world applications by providing confidence values for network predictions , Proceedings of the Sixth International Conference on Neural Information Processing (ICONIP'99), Perth .


Share

Research Platforms
N/A

Campaigns


Actions
Edit Item Edit Item