Title :
The correlation as cost function in neural networks
Author :
Englisch, H. ; Hiemstra, Y.
Author_Institution :
Inst. of Inf., Leipzig Univ., Germany
fDate :
27 Jun-2 Jul 1994
Abstract :
It is shown that in time series forecasting by feedforward neural nets with a nonlinear transfer function for the output unit the minimization of the usual cost function, the mean squared error, is not equivalent to the maximization of the correlation. A simple modification of the neural net is proposed which restores the equivalence known from linear regression
Keywords :
correlation methods; feedforward neural nets; forecasting theory; time series; transfer functions; correlation maximization; cost function minimization; feedforward neural nets; linear regression; mean squared error; nonlinear transfer function; time series forecasting; Convergence; Cost function; Economic forecasting; Error analysis; Feedforward neural networks; Intelligent networks; Linear regression; Neural networks; Robustness; Transfer functions;
Conference_Titel :
Neural Networks, 1994. IEEE World Congress on Computational Intelligence., 1994 IEEE International Conference on
Conference_Location :
Orlando, FL
Print_ISBN :
0-7803-1901-X
DOI :
10.1109/ICNN.1994.374741
