Tuning hyperparameters

In this section, we'll discover the most important hyperparameters that we can tune in BigQuery ML.

Important note

Hyperparameter tuning is the practice of choosing the best set of parameters to train a specific ML model. A hyperparameter influences and controls the learning process during the ML training stage.

By design, BigQuery ML uses default hyperparameters to train a model, but advanced users can manually change them to influence the training process.

In BigQuery ML, we can specify the hyperparameters in the OPTIONS clause as optional parameters. The most relevant hyperparameters, depending on the model, that we can change before starting the training of a BigQuery ML model are listed here:

• L1_REG: This is a regularization parameter that we can use to prevent overfitting by keeping the weights of the model close to zero.
• L2_REG: This is a second regularization parameter that we can use to prevent overfitting.
• MAX_ITERATIONS...