Automated machine learning
"Brute-force patterns finding": this is how we can briefly (and colorfully) summarize what Automated Machine Learning or, for short, AutoML, is all about. As you saw in Chapters 4 and 5, building a machine learning model is far from being a linear, single-attempt endeavor. The usual procedure for obtaining high-performing supervised models is to go through a series of "back and forth" attempts: each time, we apply some "tuning" to the model or its features and check whether the predictive performance increases or not. We have seen already some of these mechanisms in action:
- Hyperparameters optimization: this is when you apply changes to the way the learning algorithm operates, like when we activated pruning in decision trees or changed the degree of a polynomial regression. In more complex models (like in the case of deep neural networks), changing parameters (for instance, the number of neurons in the network) can make...
