DocumentCode :
969047
Title :
Strong universal consistency of neural network classifiers
Author :
Farago, Andras ; Lugosi, Gabor
Author_Institution :
Tech. Univ. of Budapest, Hungary
Volume :
39
Issue :
4
fYear :
1993
fDate :
7/1/1993 12:00:00 AM
Firstpage :
1146
Lastpage :
1151
Abstract :
In statistical pattern recognition, a classifier is called universally consistent if its error probability converges to the Bayes-risk as the size of the training data grows for all possible distributions of the random variable pair of the observation vector and its class. It is proven that if a one-layered neural network with properly chosen number of nodes is trained to minimize the empirical risk on the training data, then a universally consistent classifier results. It is shown that the exponent in the rate of convergence does not depend on the dimension if certain smoothness conditions on the distribution are satisfied. That is, this class of universally consistent classifiers does not suffer from the curse of dimensionality. A training algorithm is presented that finds the optimal set of parameters in polynomial time if the number of nodes and the space dimension is fixed and the amount of training data grows
Keywords :
error statistics; learning (artificial intelligence); neural nets; nonparametric statistics; pattern recognition; Bayes-risk; convergence rate; error probability; neural network classifiers; observation vector; one-layered neural network; polynomial time; smoothness conditions; space dimension; statistical pattern recognition; training algorithm; training data; universal consistency; Classification algorithms; Convergence; Error probability; Feedforward neural networks; Kernel; Neural networks; Pattern recognition; Polynomials; Random variables; Training data;
fLanguage :
English
Journal_Title :
Information Theory, IEEE Transactions on
Publisher :
ieee
ISSN :
0018-9448
Type :
jour
DOI :
10.1109/18.243433
Filename :
243433
Link To Document :
بازگشت