how to use neural network

i want ask how to determine epoch in matlab r2019b when the data from excel?

[Walter Roberson]

Split the task into two parts. First read the data. Then determine the epoch of the data you have in memory.

importtool and then nntool

[Cicasmi Hasibuan]

Thankyou for your answer. but i want ask again, Is Performance (as shown below) the MSE?
if not, how to find MSE ?

[Mohammad Sami]

Your model will likely become over fitted with so many epochs. The best way is to set aside data for validations and stop the training once validation accuracy starts to plateau or decreases.

[Cicasmi Hasibuan]

we have used the smallest epoch from epoch 100 - 5000. And the best result is 5000, that's why I use epoch 5000 to train tha data, and then we get the plotperform, so if the results of the plotperform are the results of MSE?

[Mohammad Sami]

Yes MSE is one of the measure you can use to check the performance.

What I was highlighting is that you are using training MSE, which can result in overfitting your model to the data. You should use validation MSE instead, which will be a better reflecting of how your model would perform.

[Cicasmi Hasibuan]

thankyou sir for your answer. so sir how to find validation MSE ? its that different from training MSE sir ?

[Abhipsa]

Hi @Cicasmi Hasibuan, actually you need to have a validation dataset in order to calculate validation MSE. Validation dataset is a separate dataset that is not used in training but can be used to help prevent overfitting. One of the possible way is "Early Stopping" i.e. to stop the model training when validation loss stops decreasing (i.e., when the model starts overfitting). We can monitor the validation MSE and stop training when it starts increasing.