0

So from what I've understood the formula of the MSE is: MSE= 1/n * ∑(t−y)^2, where n is the number of training sets, t is my target output and y my actual output. Let's say I had 2 training sets each with 1 output:
[0;0] t=[0] y=[1]

[1;1] t=[1] y=[1]

If I apply the MSE I would get MSE = 1/2 * [(0-1)^2 + (1-1)^2] = 1/2
But what if I have more than 1 output? Do I calculate the MSE of each training set and then I calculate the mean of all the MSEs I got?

L2CH
  • 13
  • 3
  • Are you sure that `n` is the number of training sets? Where have you read that? Can you provide the source of that statement? – Progman Dec 22 '19 at 13:36
  • To be fair I don't know, I just assumed that since I have many training sets, the mean of the error should be calculated by dividing the sum by the number of sets. However I don't know if this is still the case for networks with more than 1 output. – L2CH Dec 22 '19 at 15:27
  • `n` in this case is usually the number of output ports so you sum up all the errors from the output ports and build the average. I don't think this is meant as the number of training sets. – Progman Dec 22 '19 at 16:05
  • But then how would I calculate the error of the entire network? Do I take the average of all the MSEs? Like i have 2 training sets each with 2 outputs. In the first training sets the MSE is 2 and in the second set the MSE is 1, so the network's error should be (2+1)/2 = 1.5? – L2CH Dec 23 '19 at 11:56

0 Answers0