The limited long range connectivity in biological neural networks necessitates the construction of deep convolutional neural networks. In artificial neural networks, stochastic gradient descent is used to adjust parameters to solve problems based on labeled data.