Supervised learning of Perceptron and output feedback dynamic networks: A feedback analysis via the small gain theorem