In this chapter we have learned about neural networks. These are some of the most popular machine learning models used today and have been the driving force for many recent developments in the area. We have seen how neural networks can be used for both classification and regression as well as for image data. The use of neural networks for these tasks has been enabled by the abundance of labeled training data. While the success stories of neural network models are plenty, having access to such large datasets of high-quality data is still a common issue in many practical settings.

In the next section we will look at methods that do not require these large, labeled datasets. As you will see, we can sometimes learn valuable models even when no ground truth labels are available!