Title :
Sampling rate for information encoding using multilayer neural networks
Author :
Malinowski, A. ; Zurada, J.M.
Author_Institution :
Dept. of Electr. Eng., Louisville Univ., KY, USA
Abstract :
A new approach to band-limited function approximation using two-layer neural networks is presented. The Nyquist sampling rate theorem is used to solve for the optimum amount of learning data in n-dimensional input space. Choosing the least but still sufficient set of training vectors results in reduced number of hidden neurons and learning time for the network.
Keywords :
Nyquist criterion; approximation theory; encoding; feedforward neural nets; function approximation; generalisation (artificial intelligence); learning (artificial intelligence); Nyquist sampling rate theorem; band-limited function approximation; generalisation; heuristics; hidden neurons; information encoding; input space; learning data; multilayer neural networks; training vectors; Encoding; Fourier transforms; Frequency; Function approximation; Multi-layer neural network; Multidimensional systems; Neural networks; Neurons; Sampling methods; Signal restoration;
Conference_Titel :
Neural Networks, 1993. IJCNN '93-Nagoya. Proceedings of 1993 International Joint Conference on
Print_ISBN :
0-7803-1421-2
DOI :
10.1109/IJCNN.1993.716982
