Summary
The multilayer perceptron is a non-parametric estimator that has been used for several decades in various industries, from insurance to image processing. It has been proven a reliable model for classification at the cost of lengthy execution.
In this chapter, you learned the origin of feedforward neural networks, how to implement the different steps of the training cycle with backpropagation in Scala, experiment with the different configuration parameters and apply the multilayer perceptron to the analysis of the fluctuations in the currencies market.
The multilayer perceptron, along with Monte Carlo sampling (Chapter 8, Monte Carlo Inference) and regularization (Chapter 9, Regression and Regularization), is a key ingredient of deep learning, which is the topic of the next chapter.