A universal neural net with guaranteed convergence to zero systemerror
Chang, T.-S.
Abdel-Ghaffar, K.A.S.
Dept. of Electr. Eng. & Comput. Sci., California Univ., Davis, CA;
This paper appears in: Signal Processing, IEEE Transactions on
Publication Date: Dec 1992
Volume: 40,
Issue: 12
On page(s): 3022-3031
ISSN: 1053-587X
References Cited: 8
CODEN: ITPRED
INSPEC Accession Number: 4346416
Digital Object Identifier: 10.1109/78.175745
Current Version Published: 2002-08-06
Abstract
A learning algorithm with guaranteed convergence to zero system
error is developed. The algorithm also has high potential to converge
fast. The basic idea is to let the net grow when stopping at a local
minimum, so that the original local minimum is no longer a local minimum
with regard to the new net, and the new net always starts from a point
with less error than that in the original local minimum. By this method,
the error is guaranteed to decrease until it converges to zero. The
technique can also be used to make the error to be as small as desired
when the backpropagation algorithm reaches a global minimum which does
not achieve zero error due to a lack of sufficient number of nodes. When
expanding the neural net, the initial weights of the new node can be
selected to maximize the error gradient. A mathematical proof of the
guaranteed learning of the universal neural net is given, and numerical
examples illustrate its high potential for fast learning
Index
Terms
Available to subscribers and IEEE members.
References
Available to subscribers and IEEE members.
Citing Documents
Available to subscribers and IEEE members.
Cart
