I want to program NN with many outputs, how many outputs is limit for NN? if output number is large, what is the relationship between traning data and outputs?
Also, your question is quite difficult to understand, and very inexact.
Basically, you can have an arbitrary number of outputs for your neuronal net, you are only limited by memory and processor speed.
If, by output, you mean what classification can be done with a given neural net, this also is solvable only if you give the appropriate activation function, and the depth of the net.
Generally, for the relation between training data and outputs, the larger your neuronal net is and the more clusters you want to seperate, the more training data is required.
Thank Wernaeh for your information!