Machine Learning in Elixir

Chapter 7Learn to See

In the previous chapter, you used Axon to create and train neural networks in Elixir, including one that recognizes handwritten digits. You also learned about the types of problems neural networks are well-suited to handle and why they outperform shallow machine learning approaches in many areas. More specifically, you implemented a type of neural network architecture known as the multi-layer perceptron (MLP) or deep feed-forward network.

While MLPs are capable of learning and modeling any kind of data, other deep learning architectures are even more powerful when applied to specific problem types. In this chapter, you’ll implement domain-specific architecture known as the convolutional neural network (CNN) and learn how ...