Title :
Incremental Learning Framework for Function Approximation via Combining Mixture of Expert Model and Adaptive Resonance Theory
Author :
Kim, Cheoltaek ; Lee, Ju-Jang
Author_Institution :
Korea Adv. Inst. of Sci. & Technol., Daejeon
Abstract :
This paper introduces an incremental learning framework for function approximation which uses the structure of mixture of expert model and learning methodology of adaptive resonance theory. The proposed framework adapts their structure and parameter values through incremental, competitive learning, and supervised learning. The main idea comes from that the combination of two classical methods which are mixture of expert model and adaptive resonance theory can be jointly learned and the combination keeps up the advantages of each method;the mixture of expert model has the ability to avoid strong interference and the adaptive resonance theory is one of the best model of incremental learning. The idea can be implemented by modifying adaptive resonance theory based on the mixture of expert model. The empirical experiment would show the performance of the proposed implementation via comparing receptive field weighted regression(RFWR) and PROBART.
Keywords :
adaptive resonance theory; expert systems; function approximation; mathematics computing; regression analysis; unsupervised learning; adaptive resonance theory; competitive learning; expert model; function approximation; incremental learning; receptive field weighted regression; supervised learning; Automation; Computer science; Convergence; Function approximation; Interference; Mechatronics; Resonance; Shape; Subspace constraints; Supervised learning; Adaptive Resonance Theory; Function Approximation; Incremental Learning; Mixture of Experts; Multilayer Perceptron;
Conference_Titel :
Mechatronics and Automation, 2007. ICMA 2007. International Conference on
Conference_Location :
Harbin
Print_ISBN :
978-1-4244-0828-3
Electronic_ISBN :
978-1-4244-0828-3
DOI :
10.1109/ICMA.2007.4304124