In this chapter, you learned how to solve a handwritten digit-recognition problem. Starting from the basis of the OCR and computer vision concepts, we learned how to elaborate simple images.
Then, an autoencoder was used for handwritten digit recognition. An autoencoder is a neural network whose purpose is to code its input into small dimensions and the result obtained to be able to reconstruct the input itself. The purpose of autoencoders is not simply to perform a sort of compression of the input or look for an approximation of the identity function; but there are techniques that allow us to direct the model (starting from a hidden layer of reduced dimensions) to give greater importance to some data properties. Thus they give rise to different representations based on the same data.
Finally, autoencoders and reinforcement learning concepts were joined to improve the...