rms.m

人工神经网络：MATLAB源程序用于训练测试

%# function [RMSE,PMSEtest,PRESS,SE,BIAS,SE2] = rms(y,yhat)						
%#											
%# AIM:		Calculates Root Mean Squared Error (RMSE), PRedicted REsidual Sum	
%#		of Squared errors (PRESS), Standard Error (SE) and bias for a vector 	
%#		of predicted y-values.							
%#											
%# INPUT:	y : n-element vector of experimental responses (row- or column-vector)	
%#		yhat : n-element vector of estimated responses (same dimentions of y)	
%#											
%# OUTPUT:	RMSE : Root Mean Squared Error						
%#		PRESS : PRedicted REsidual Sum of Squared error				
%#		SE : Standard Error							
%#		BIAS : bias								
%#											
%# AUTHOR:	Frederic Despagne							
%#		Copyright(c) 1997 for ChemoAC						
%#		Dienst FABI, Vrije Universiteit Brussel					
%#		Laarbeeklaan 103, 1090 Jette						
%#										
%# VERSION: 1.1 (28/02/1998)								
%#											
%# TEST:	Roy De Maesschalck							

function [RMSE,RMStest,PRESS,SE,BIAS,SE2] = rms(y,yhat);

n = length(y);				% Number of samples
resid = yhat-y;				% Vector of residuals
SE = sqrt(cov(resid));			% SE


BIAS = (sum(resid))/n;			% Bias
PRESS = sum((resid).^2);		% PRESS
RMSE = sqrt(PRESS/n);			% RMSE

RMStest=sqrt(resid'*resid/n);
SE2= sqrt(sum((yhat-y-BIAS).^2)/(n-1));