Title :
On optimal radial basis function nets and nonlinear function estimates
Author_Institution :
Dept. of Comput. Sci., Concordia Univ., Montreal, Que.
Abstract :
Radial basis function (RBF) networks with one hidden layer are considered. Using the connections between RBF nets and the kernel regression estimates (KRE) upper bounds on L2 errors of RBF nets are derived and optimized with respect to the radial functions. Analytical expressions the optimal radial functions are given and the optimal rates of convergence in the class smooth functions are derived
Keywords :
estimation theory; feedforward neural nets; learning (artificial intelligence); statistical analysis; kernel regression estimates; nonlinear function estimates; optimal radial basis function nets; optimal rates of convergence; smooth functions; Computational Intelligence Society; Computer errors; Computer science; Convergence; Kernel; Neural networks; Radial basis function networks; Regression analysis; Tail; Upper bound;
Conference_Titel :
Neural Networks, 1995. Proceedings., IEEE International Conference on
Conference_Location :
Perth, WA
Print_ISBN :
0-7803-2768-3
DOI :
10.1109/ICNN.1995.488106
