MATLAB: Regression Plot

regression

As regards the neural Network Training, in the function I set a percent of data to be sued for testing and an another amount for validation. After training we can see the regression plot for all the data set. This regression plot which is launched by clicking on the Regression on the training window is much better than the results obtained from separate data used to test the network….!!! it's the matter every time. Always it' better than the result I obtain when using other bunch of data.

Your help is appreciated.

Best Answer

No mystery.

The training algorithm tries to minimize the error on the training subset. That is why the training error is, most of the time, lower than that of nontraining data.

Related Solutions

MATLAB: Is the mse of the testing set in the neural network computed simultaneously with the mse of the training and validation sets

Yes.

Thank you for formally accepting my answer

Greg

MATLAB: Regression

Here is an example:

x = (1:20)';
y = 2 + 3*x + 7*rand(20,1);
model = fit(x,y,'poly1')
figure
plot(model,x,y)

The plot command here is an "overloaded" method for cfit objects (the output of the fit command). The documentation is here: http://www.mathworks.com/help/toolbox/curvefit/plot.html

Best Answer

Related Solutions

MATLAB: Is the mse of the testing set in the neural network computed simultaneously with the mse of the training and validation sets

MATLAB: Regression

Related Question