Introduction

In the previous chapter, we learned about pre-trained networks and how to utilize them for our own applications via transfer learning. We experimented with VGG16 and ResNet50, two pre-trained networks that are used for image classification, and used them to classify new images and fine-tune them for our own applications. By utilizing pre-trained networks, we were able to train more accurate models quicker than the convolutional neural networks we trained in previous chapters.

In traditional neural networks (and every neural network architecture covered in prior chapters), data passes sequentially through the network from the input layer, and through the hidden layers (if any), to the output layer. Information passes through the network once and the outputs are considered independent of each other, and only dependent on the inputs to the model. However, there are instances where a particular output is dependent on the previous output of the system.

Consider the stock...