when you say output models are you referring to the optimized ANNs, or the organisation of output neurons?