Problem with Cross Validation Neural Network

Hi, I want to build a Cross-Validation neural network for a regression problem with trainlm as the training function and mse as the performance. I have run the network for more than 10 times but each time the errors for each step of K was 0 except the last step! For example when K=5, test errors will be :

error for K=1: 0

error for K=2: 0

error for K=3: 0

error for K=4: 0

error for K=5: "A large value"

My question is why only error is large for the last iteration of Cross-Validation? I would be so grateful if someone help me to know what is wrong with this network !. Thanks in advance.

[Greg Heath]

Have you double-checked equations 5 through 12?

[Mohammadreza Hesami]

Thanks Greg for your reply. Unfortunately, I can't get your point, could please tell me what do you mean by these equations or where can I find them ? Thanks in advance