Deep learning is a subject of importance right from image detection to speech recognition and AI-related activity. There are numerous products and packages in the market for deep learning. Some of these are Keras, TensorFlow, h2o, and many others.

In this chapter, we learned the basics of deep learning, many variations of DNNs, the most important deep learning algorithms, and the basic workflow for deep learning. We explored the different packages available in R to handle DNNs.

To understand how to build and train a DNN, we analyzed a practical example of DNN implementation with the neuralnet package. We learned how to normalize data across the various available techniques, to remove data units, allowing you to easily compare data from different locations. We saw how to split the data for the training and testing of the network. We learned to use the neuralnet function to build and train a multilayered neural network. So we understood how to use the trained network to make predictions and we learned to use the confusion matrix to evaluate model performance.

We saw some basics of the h2o package. Overall, The h2o package is a highly user-friendly package that can be used to train feed-forward networks or deep autoencoders. It supports distributed computations and provides a web interface. By including the h2o package like any other package in R, we can do all kinds of modeling and processing of DNNs. The power of h2o can be utilized by the various features available in the package.

In the next chapter, we will understand what a perceptron is and what are the applications that can be built using the basic perceptron. We will learn a simple perceptron implementation function in R environment. We will also learn how to train and model a MLP . We will discover the linear separable classifier.