Title :
Modified stochastic approximation to enhance unsupervised learning
Author :
Schwartz, S.C. ; Katopis, A.
Author_Institution :
Princeton University, Princeton, New Jersey
Abstract :
By simple modifications of a decision-directed learning procedure, the regression curves of multidimensional stochastic approximation can be rotated further apart, leading to enhanced convergence properties. Results of a Monte Carlo simulation for a binary hypotheses testing problem are given which illustrates this faster convergence.
Keywords :
Computer science; Convergence; Equations; Gaussian noise; Jacobian matrices; Multidimensional systems; Pattern recognition; Stochastic processes; Testing; Unsupervised learning;
Conference_Titel :
Decision and Control including the 16th Symposium on Adaptive Processes and A Special Symposium on Fuzzy Set Theory and Applications, 1977 IEEE Conference on
Conference_Location :
New Orleans, LA, USA
DOI :
10.1109/CDC.1977.271728
