Single layer neuron network is trained for simple set of mappings:
http://www.wolfram.com/cdf-player/ <---- download this free plugging to see the calculations
The emphasis was to make the NN training more accessible for programmers without requiring a PhD.
The code was actually implemented in Mathematica 8.0 but the pseudo C code is quite accurate.
The algorithm was matched to number of papers and 8000 training loop was experimented with.
Ok now I have the quad copter simulator, primitive but mathematically accurate and programmable. Also have the training algorithm above for single layer NN. These are Mathematica coded, but I can easily convert them to python or C if required.
So now we can experiment with NN control systems for adrucopter by off-line and in-flight training.
My next effort is to come up with NN training for vertical lift-off of the arducopter autonomously.