Fig. 4From: Superiority of quadratic over conventional neural networks for classification of gaussian mixture data BP process to train a neural network of L layersBack to article page