The success of deep learning has immensely facilitated the work of feature engineering. Indeed, traditional machine learning depended very much on the selection of the right set of features, and very frequently, this step was more important that the selection of a particular learning algorithm. Deep learning has changed this scenario; creating a right model is still very important but nowadays networks are less sensitive to the selection of a particular set of feature and are much more able to auto-select the features that really matter.

Instead, the introduction of deep learning has increased the focus on the selection of the right neural network architecture. This means that progressively the interest of researchers is shifting from feature engineering to network engineering. AutoML (Meta Learning) is an emerging research topic which aims at auto-selecting the most efficient neural network for a given learning task. In other words, AutoML represents a set of methodologies for learning how to learn efficiently. Consider for instance the tasks of Machine Translation, or Image Recognition, or Game playing. Typically, the models are manually designed by a team of engineers, data scientist, and domain experts. If you consider that a typical 10-layer network can have ~10¹⁰ candidate network, you understand how expensive, error prone, and ultimately sub-optimal the process can be.