1
$\begingroup$

If anyone has used the Neural Network toolbox in matlab, what does the two values in the performance section means? Like considering this image:
enter image description here
The performance section has the value 0.484 and 1.32e-07. what are these two values. Among them which is the mean squared error value. If one is the mean squared error value, then what is the other value?

$\endgroup$
1
$\begingroup$

I don't see it explicitly discussed in the documentation, but it appears that the numbers on the right side of the screen are the stopping criteria. That run appears to have stopped because the error on the validation set increased six times (explained here). I'm guessing it would also stop after 1000 epochs, when 0 MSE was achieved, the gradient was below 1e-6, or after 4 resets.

Thus the number to the left of the bar should be the MSE on the epoch with the best validation error*. I'm guessing the value of the blue bar is the lowest that MSE got during the iterations that were actually increasing the validation error.

From the docs: "In the default setup, the training stops after six consecutive increases in validation error, and the best performance is taken from the epoch with the lowest validation error."

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.