what is the accepted MSE could I use in training ANN?

I'm trying to train ANN with ABC algorithm the stop condition I use is 1000 epochs but it take too long time so I decided to use a good MSE value to stop training what should this value be????

[John D'Errico]

That is impossible to answer. The acceptable MSE is what might be acceptable to you. Only you know how much noise you might expect in the system. You do not want to overfit any model, as then you are just chasing noise. There is no purpose in a model that predicts the noise well. Of course, this is why you have validation and training sets. But nobody here can ever say (for example) that 0.01 is acceptable for an MSE, but 0.1 is too high. We don't know what your data means, where it comes from, what noise corrupts it, etc.

[Samisam]

hmmmmmmm how then can I compare two algorithms what do you suggest ??? I wish I could compare them with MSE value by the way the data set is Iris data set

[Walter Roberson]

You can compare the MSE of two different predicted outputs (though for sufficiently large models you could run into precision problems.) What we cannot tell you is any particular absolute MSE value to stop training at.

You are more likely to want to stop training when the training has stopped producing improvements for some number of iterations.