Reduce number of epochs for production runs

[dwr-ycheng]

Currently in this experimental phase, the number of epochs has been set to 40000 for EC stations and 30000 for X2. This makes the ANN training time in the Python script longer than that of MATLAB script. And based on the Loss Over Epochs plot, the high number of epochs will not be necessary anymore after the experimental runs.

For upcoming release of ANN in Python script (for ANN DLL production run) in Github, we should find out and use the optimal/realistic epochs number.