Title :
On hidden nodes for neural nets
Author :
Mirchandani, Gagan ; Cao, Wei
Author_Institution :
Dept. of Comput. Sci. & Electr. Eng., Vermont Univ., Burlington, VT, USA
fDate :
5/1/1989 12:00:00 AM
Abstract :
Recent results indicate that the number of hidden nodes (H) in a feedforward neural net depend only on the number of input training patterns (T). There appear to be conjectures that H is on the order of T-1 and of log2 T. A proof is given that the maximum number of separable regions (M) in the input space is a function of both H and input space dimension (d). The authors also show that H =M -1 and H=log2M are special cases of that formulation. M defines a lower bound on T, the number of input patterns that may be used for training. Application to some experiments are investigated
Keywords :
computer graphics; computerised picture processing; neural nets; experiments; feedforward neural net; hidden nodes for neural nets; input space dimension; input training patterns; maximum number of separable regions; multilayered networks; Circuits and systems; Computer science; Feedforward neural networks; Multilayer perceptrons; Neural networks; Pattern classification; Random number generation; Shape; Sonar;
Journal_Title :
Circuits and Systems, IEEE Transactions on
