Customizing hyperparameter tuning
In this section, we’ll see how AutoML can be used to tune a specific algorithm’s hyperparameters by following a search algorithm, to iteratively try different values for those hyperparameters.
If we consider each hyperparameter that a model trainer has, we can see that a hyperparameter typically has a range of valid values from some minimum amount to some maximum amount. You can visualize this as a slider on a user interface where hyperparameter tuning involves trying to find the best value for a setting.
Most model trainers have multiple hyperparameters, so you’re not just searching one hyperparameter for the best value; instead, you’re searching for an optimal set of values between your different hyperparameters.
Understanding the search space
This set of possible combinations of values is referred to as a search space, where your job as a data scientist is to find the right combinations of hyperparameter values...
