Understanding random search

Random search is the third and the last hyperparameter-tuning method that belongs in the exhaustive search group. It is a simple method but works surprisingly well in practice. As implied by its name, random search works by randomly selecting hyperparameter values in each iteration. There's nothing more to it. The selected set of hyperparameters in the previous iteration will not impact how the method selects another set of hyperparameters in the following iterations. That's why random search is also categorized as an uninformed search method.

You can see an illustration of the random search method in the following diagram:

Figure 3.4 – Random search illustration

Random search usually works better than grid search when we have little or no idea of the proper hyperparameter space for our case, and this applies most of the time. Compared to grid search, random search is also more efficient in terms of computing...