In the previous chapter, we discussed neural networks and their basic operation. Specifically, we discussed the fully connected feedforward neural network, which is just one simple topology out of many possible ANN topologies. In this chapter, we're going to focus on two advanced topologies: the Convolutional Neural Network (CNN) and one form of recurrent neural network (RNN), called the Long Short-Term Memory (LSTM) network. CNNs are used most often for image processing tasks, such as object detection and image classification. LSTM networks are often used in NLP or language-modeling problems.
These exotic ANN topologies are considered to be deep neural networks (DNNs). While the term is not well-defined, DNNs are typically understood to be ANNs with multiple hidden layers between the input and output layers. Convolutional network architectures can become...