Deep Learning Using Apache Spark

In this chapter, we will go on a hands-on exploration of the exciting and cutting-edge world of deep learning! We will use third-party deep learning libraries in conjunction with Apache Spark's MLlib to perform accurate optical character recognition (OCR) and automatically recognize and classify images via the following types of artificial neural networks and machine learning algorithms:

• Multilayer perceptrons
• Convolutional neural networks
• Transfer learning

Artificial neural networks

As we studied in Chapter 3, Artificial Intelligence and Machine Learning, an artificial neural network (ANN) is a connected group of artificial neurons that is aggregated into three types of linked neural layers—the input layer, zero or more hidden layers, and the output layer. A monolayer ANN consists of just one layer of links between the input nodes and output nodes, while multilayer ANNs are characterized by the segmentation of artificial neurons across multiple linked layers.

An ANN where signals are propagated in one direction only—that is, the signals are received by the input layer and forwarded to the next layer for processing—are called feedforward networks. ANNs where a signal may be propagated back to artificial neurons or neural layers that have already processed that signal are called feedback networks.

Backwards propagation...

Summary

In this chapter, we went on a hands-on exploration through the exciting and cutting-edge world of deep learning. We developed applications to recognize and classify objects in images with astonishingly high rates of accuracy, and demonstrated the truly impressive learning ability of ANNs to detect and learn patterns in input data.

In the next chapter, we will extend our deployment of machine learning models beyond batch processing in order to learn from data and make predictions in real time!