Linear models are found everywhere. Their simplicity, as well as the capabilities they offer—such as regularization—makes them popular among practitioners. They also share many concepts with neural networks, which means that understanding them will help you in later chapters.
Being linear isn't usually a limiting factor as long as we can get creative with our feature transformation. Furthermore, in higher dimensions, the linearity assumption may hold more often than we think. That's why it is advised to always start with a linear model and then decide whether you need to go for a more advanced model.
Having that said, it can sometimesbe tricky to figure out the best configurations for your linear model or decide on which solver to use. In this chapter, we learned about using cross-validation to fine-tune a model's hyperparameters. We have also seen the different hyperparameters and solvers available, with tips for when to use...
