Regularization of Evolving Polynomial Models. Pavel Kordik. IWIM, Prague, 2007.
Article (in pdf)
Abstract. Black box models such as neural networks are popular because they can deliver reasonably accurate model almost instantly. Sometimes, it is more convenient to use a math model instead of black box model. Math models can be either designed by experts or automatically generated from data describing modelled systems. The disadvantage of generated math models is that they are often too complex to be understood by experts. In this contribution we experiment with regularization of generated models to enable automatic evolution of models that are both enough accurate and understandable. We limit our experiments to models consisting of polynomial transfer units.
Keywords. Inductive modelling, regularization, polynomial networks.
References.
- P. Bearse and H. Bozdogan. Subset selection in vector autoregressive models using the genetic algorithm with informational complexity as the fitness function. Systems Analysis, Modelling, and Simulation (SAMS), 31:61–91, 1998.
- V. P. Belogurov. A criterion of model suitabilityfor forcasting quantitative processes. Soviet Journal of Automation and Information Sciences, 23(3):21–25, 1990.
- A. Ivakhnenko, E. Savchenko, and G. Ivakhnenko. Gmdh algorithm for optimal model choice by the external error criterion with the extension of definition by model bias and its applications to the committees and neural networks. Pattern Recognition and Image Analysis, 12(4):347353, 2002.
- A. G. Ivakhnenko, G. Ivakhnenko, and J. Muller. Self-organization of neural networks with active neurons. Pattern Recognition and Image Analysis, 4(2):185–196, 1997.
- P. Kord´ık. Fully Automated Knowledge Extraction using Group of Adaptive Models Evolution. PhD thesis, Czech Technical University in Prague, FEE, Dep. of Comp. Sci. and Computers, FEE, CTU Prague, Czech Republic, September 2006.
- H. Madala and A. Ivakhnenko. Inductive Learning Algorithm for Complex System Modelling. CRC Press, 1994. Boca Raton.
- K. O. Stanley and R. Miikkulainen. Evolving neural networks through augmenting topologies. Evolutionary Computation, 10(2):99–127, 2002. Massachusetts Institute of Technology.
