Hyperparameter Tuning
Umberto Michelucci · Apress eBooks · 2018
In this chapter, you will look at the problem of finding the best hyperparameters to get the best results from your models. First, I will describe what a black-box optimization problem is, and how that class of problems relate to hyperparameter tuning. You will see the three best-known methods to tackle these kind of problems: grid search, random search, and Bayesian optimization. I will show you, with examples, which one works under which conditions, and I will give you a few tricks that are very helpful for improving optimization and sampling on a logarithmic scale. At the end of the chapter, I will show you how you can use those techniques to tune a deep model, using the Zalando dataset.