Patent
1991-12-27
1994-01-04
Fleming, Michael R.
395 11, 395 22, 395 25, G06F 1516
Patent
active
052767710
ABSTRACT:
A data processing system and method for solving pattern classification problems and function-fitting problems includes a neural network in which N-dimensional input vectors are augmented with at least one element to form an N+j-dimensional projected input vector, whose magnitude is then preferably normalized to lie on the surface of a hypersphere. Weight vectors of at least a lowest intermediate layer of network nodes are preferably also constrained to lie on the N+j-dimensional surface.
To train the network, the system compares network output values with known goal vectors, and an error function (which depends on all weights and threshold values of the intermediate and output nodes) is then minimized. In order to decrease the network's learning time even further, the weight vectors for the intermediate nodes are initially preferably set equal to known prototypes for the various classes of input vectors. Furthermore, the invention also allows separation of the network into sub-networks, which are then trained individually and later recombined. The network is able to use both hyperspheres and hyperplanes to form decision boundaries, and, indeed, can converge to the one even if it initially assumes the other.
REFERENCES:
patent: 3097349 (1963-07-01), Putzrath et al.
patent: 3158840 (1964-11-01), Baskin
patent: 3209328 (1965-09-01), Bonner
patent: 3230351 (1966-01-01), Platt et al.
patent: 3275986 (1966-09-01), Dunn et al.
patent: 3280257 (1966-10-01), Orthuber et al.
patent: 3284772 (1966-11-01), Stewart et al.
patent: 3284780 (1966-11-01), Clapper
patent: 3310783 (1967-03-01), Putzrath
patent: 3310784 (1967-03-01), Hilinski
patent: 3311895 (1967-03-01), Clapper
patent: 3317900 (1967-05-01), Clapper
patent: 3317901 (1967-05-01), Clapper
patent: 3324457 (1967-06-01), Ogle et al.
patent: 3325787 (1967-06-01), Angell et al.
patent: 3333248 (1967-07-01), Greenberg et al.
patent: 3333249 (1967-07-01), Clapper
patent: 3351783 (1967-11-01), Harris et al.
patent: 3408627 (1968-10-01), Kettler et al.
patent: 3435422 (1969-03-01), Gerhardt et al.
patent: 3440617 (1969-04-01), Lesti
patent: 3533072 (1970-10-01), Clapper
patent: 3548202 (1970-12-01), Ide et al.
patent: 3602888 (1971-08-01), Nishiyama et al.
patent: 3701974 (1972-10-01), Russell
patent: 4163983 (1979-08-01), Cline et al.
patent: 4254474 (1981-03-01), Cooper et al.
patent: 4326259 (1982-04-01), Cooper et al.
patent: 4450530 (1984-05-01), Llinas et al.
patent: 4518866 (1985-05-01), Clymer
patent: 4719591 (1988-01-01), Hopfield et al.
patent: 4730259 (1988-03-01), Gallant
patent: 4731747 (1988-03-01), Denker
patent: 4737929 (1988-04-01), Denker
patent: 4752906 (1988-06-01), Kleinfeld
patent: 4760437 (1988-07-01), Denker et al.
patent: 4760604 (1988-07-01), Cooper et al.
patent: 4766568 (1988-08-01), McGregor et al.
patent: 4774677 (1988-09-01), Buckley
patent: 4796199 (1989-01-01), Hammerstrom et al.
patent: 4800519 (1989-01-01), Grinberg et al.
patent: 4803736 (1989-02-01), Grossberg et al.
patent: 4805225 (1989-02-01), Clark
patent: 4807168 (1989-02-01), Moopenn et al.
patent: 4979126 (1990-12-01), Pao et al.
patent: 5014219 (1991-05-01), White
patent: 5105468 (1992-04-01), Guyon et al.
patent: 5121231 (1992-06-01), Jenkins et al.
patent: 5132811 (1992-07-01), Iwaki et al.
patent: 5170463 (1992-12-01), Fujimoto et al.
patent: 5179596 (1993-01-01), Weingard
Teuvo Kohonen, "Self-Organization and Associative Memory," (Berlin: Springer-Verlag 1984).
James A. Anderson and Edward Rosenfeld, Ed., "Neurocomputing Foundations of Research," (London: The MIT Press 1988).
Richard P. Lippman, "An Introduction to Computing with Neural Nets," IEEE ASSP Magazine, Apr. 1987, pp. 4-22.
Gail A. Carpenter, Stephen Grossberg, and David Rosen, Center for Adaptive Systems and Graduate Program in Cognitive & Neural Systems, Boston University, "ART 2-A: An Adaptive Resonance Algorithm for Rapid Category Learning and Recognition," Published by IEEE, 1991.
Gail A. Carpenter, Stephen Grossberg, and David Rosen, Center for Adaptive Systems and Graduate Program in Cognitive & Neural Systems, Boston University, "Fuzzy Art: An Adaptive Resonance Algorithm for Rapid, Stable Classification of Analog Patterns," Published by IEEE, 1991.
Teuvo Kohonen, "Self-organized Formation of Topologically Correct Feature Maps," Biological Cybernetics 43: 59-69, Springer-Verlag 1982.
Teuvo Kohonen, "Correlation Matrix Memories," IEEE Transactions on Computers C-21: 353-359, 1972.
D. E. Rumelhardt, G. E. Hinton, and R. J. Williams, "Learning Internal Representations By Error Propagation," Parallel Distributed Processing: Explorations in the Microstructures of Cognition, vol. 1, D. E. Rumelhart and J. L. McClelland (Eds.) (Cambridge, Mass.: MIT Press 1986), pp. 318-362.
Douglas L. Reilly, Leon N. Cooper, and Charles Elbaum, "A Neural Model for Category Learning," Biological Cybernetics 45, 35-41, Springer-Verlag 1982.
Amir Sarajedini, Robert Hecht-Nielsen, University of California, San Diego, "The Best of Both Worlds: Casasent Networks Integrate Multilayer Perceptrons and Radial Basis Functions." Paper submitted for publication to IJCNN, 1992, Baltimore.
James Saffery and Chris Thornton, Department of Artifical Intelligence, Edinburgh, "Using Stereographic Projection as a Preprocessing Technique for Upstart," published by IEEE, Jan. 15, 1991.
Brian Telfer and David Casasent, Center for Excellence in Optical Data Processing, Carnegie Mellon University, "Minimum-Cost Ho-Kashyap Associative Processor for Piecewise-Hyperspherical Classification," published by IEEE, 1991.
Manukian Narbik
Wilensky Gregg D.
Fleming Michael R.
Hafiz Tarig R.
R & D Associates
LandOfFree
Rapidly converging projective neural network does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Rapidly converging projective neural network, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Rapidly converging projective neural network will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-313931