Art-Based Autonomous Learning Systems: Part I — Architectures and Algorithms
This chapter describes the design of novel ART-based intelligent systems that are able to learn and, at the same time, to refine their knowledge in perpetuity. Fuzzy ARTMAP and the Probabilistic Neural Network are integrated to form a hybrid system that possesses the desirable properties for incremental, causal learning as well as for Bayesian probability estimation. Subsequently, a multiple neural network architecture is devised to aggregate outputs from several individual networks into a unified decision. A number of algorithms is proposed to increase the generalization and adaptability of the resulting systems.
KeywordsInput Pattern Probabilistic Neural Network Target Class Belief Function Adaptive Resonance Theory
Unable to display preview. Download preview PDF.
- Rumelhart, D.E, Hinton, G.E., and Williams, RJ. (1986), “Learning internal representation by error propagation,” in Rumelhart, D.E. and McLelland, J.L. (Eds.), Parallel Distributed Processing, I, pp. 318–362, Cambridge, MA: MIT Press.Google Scholar
- Specht, D.F. (1990), “Probabilistic neural networks,” Neural Networks, Vol. 3, pp. 109–118.Google Scholar
- Fukunaga, K. (1972), Introduction to Statistical Pattern Recognition, New York: Academic Press.Google Scholar
- Masters, T. (1993), Practical Neural Network Recipes in C++, San Diego, CA: Academic Press.Google Scholar
- Masters, T. (1995), Advanced Algorithms for Neural Networks: A C++ Sourcebook, New York: John Wiley and Sons.Google Scholar
- Pearl, J. (1988), Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference, San Mateo, CA.: Morgan Kaufmann.Google Scholar