The number of epochs depends on your problem. You should use enough epochs to avoid over fitting. in the problems that I have seen , it was less than 200.

The basic idea behind using a checking data set for model validation is that after a certain point in the training, the model begins overfitting the training data set. In principle, the model error for the checking data set tends to decrease as the training takes place up to the point that overfitting begins and then the model error for the checking data suddenly increases.

for further information, you can refer to Matlab FL toolbox manuals and related articles. There are many publications in this regard including mine:

http://www.researchgate.net/publication/257389427_Prediction_of_elastic_modulus_of_normal_and_high_strength_concrete_using_ANFIS_and_optimal_nonlinear_regression_models

Article Prediction of elastic modulus of normal and high strength co...

There is no particular optimum. for two reasons

1) The optimisation function of ANFIS is basically gradient descent. These algorithms either reach optimum quickly (few tens of epochs max) or never.

2) Real data comprise noise so you expect some prediction error whatsoever.

3) you can create your own stop criterion by inserting som threshold in objective function value change or prediction error value change assuming a validation set.

Try to change from 30 to 50 to 100 to 1000 and see that the differences are not significant (I predict).

Personally I choose the matlb default and it produces adequate answers.

The number of epochs depends on your problem. You should use enough epochs to avoid over fitting. in the problems that I have seen , it was less than 200.

The basic idea behind using a checking data set for model validation is that after a certain point in the training, the model begins overfitting the training data set. In principle, the model error for the checking data set tends to decrease as the training takes place up to the point that overfitting begins and then the model error for the checking data suddenly increases.

for further information, you can refer to Matlab FL toolbox manuals and related articles. There are many publications in this regard including mine:

http://www.researchgate.net/publication/257389427_Prediction_of_elastic_modulus_of_normal_and_high_strength_concrete_using_ANFIS_and_optimal_nonlinear_regression_models

Article Prediction of elastic modulus of normal and high strength co...