Abstract
This paper focuses on the evaluation of learning performance in intelligent dynamic processes with supervised learning. Learning dynamics are characterized by basins of attraction generated by state transitions in control space (statespace + parameter space). State uncertainty is modelled as a cellular control space, namely the cell space. Learning performance losses are related to nonseparable basins of attractions with fuzzy boundaries and to their erosions under parameter changes. Basins erosions are analyzed as fingering regions which quickly loose their compactness yielding regions of fractional dimensions and degeneracies due to bifurcation phenomena. We therefore claim that “learning” quality of intelligent dynamic processes should be measured by fractal set theoridc methods.
To this end, we generate in this paper learning patterns as convergence maps using the cell to cell mapping concept. We then evaluate predictability of these patterns based on Lyapunov exponents. Performance measures in training are generated based on box counting fractal dimensions and the lose of reliability is detected by bifurcation phenomena. Illustrative results are reported for a collision free intelligent path planner of a planar robot manipulator.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
A.M. Erkmen, F. Yeğenoğlu, and H.E. Stephanou, ‘Entropy-Driven On-Line Control of Automonous Robots’, International Journal of Intelligent and Robotics Systems, special issue on ‘Intelligent Control’, Vol. 2, Nos 2-3, 1989.
A.M. Erkmen, ‘A Conductivity-Based Evidential Classifier for the Recognition of Fractal Patterns in Robot Sensing’, Proc. of Bilkent International Conference on New Trends in Communication, Control and Signal Processing (IEEE, EURASIP), 1990.
M. Üneri and A.M. Erkmen, ‘Generating Performance Criteria for Sensory Motor Trainer Sets of a Robot Manipulator using Cell-to-Cell Mapping’, Proc. of IEEE/RSJ (Robotic Society of Japan) International Conference on Intelligent Robots and Systems, 1992.
M. Üneri and A.M. Erkmen,‘Bifurcation Phenomena in Sensorimotor training of a Robot Manipulator’, IFAC (International Federation of Automatic Control) preprints of the International Workshop on Automatic Control for Quality and Productivity, 1992.
M. Oneri and A.M. Erkmen, ‘Erosion of Basins of Attraction: Performance Losses in Sensorimotor Learning of a Robot Manipulator’, proc. IEEE International Symposium of Intelligent Control, 1992.
C.S. Hsu, ‘A Theory of Cell-to-Cell Mapping Dynamical Systems’, J. Appl. Mechanics, Vol.47, Dec. 1980.
C.S. Hsu Cell-to-Cell Mapping, New York: Splinger-Verlag, 1987.
C.S. Hsu and R.S. Guttalu, ‘An Unravelling Algorithm for Global Analysis of Dynamical Systems: An Application of Cell-to-Cell Mappings’, J. Appl. Mechanics, Vol.47,Dec. 1980.
E.J. Kreuzer, ‘Domains of Attraction in Systems with Limit Cycles’, in Proc. of German-Japanese Seminar on Nonlinear Problems in Dynamical Systems, Universitat Stutgart 1984.
J.M.T.T. Thomson, F.R.S. and M.S. Soliman,‘Fractal Control Boundaries of Driven Oscillators and their Relevance to Safe Engineering Design’, Proc. R. Soc. Lond., A 428, 1990.
J. Guckenheimer, P. Holmes, Nonlinear Oscilations Dynamical Systems and Bifurcation of Vector Fields, Splinger Verlag, 1983.
K.J. Falconer, The Geometry of Fractal Sets, Cambridge University Press, 1985.
K.J. Falconer,‘Fractal Geometry: Mathematical Foundation and Application’, John Viley & Sons, NY, 1990.
P. Hagedorn,Nonlinear Oscillations, Clarendon Press, Oxford, 1988.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 1993 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Erkmen, A.M. (1993). Fractal Set Theoretic Analysis of Performance Losses for Tuning Training Data in Learning Systems. In: Kaynak, O., Honderd, G., Grant, E. (eds) Intelligent Systems: Safety, Reliability and Maintainability Issues. NATO ASI Series, vol 114. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-58021-5_11
Download citation
DOI: https://doi.org/10.1007/978-3-642-58021-5_11
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-63438-3
Online ISBN: 978-3-642-58021-5
eBook Packages: Springer Book Archive