What are hyperparameters in the context of machine learning?

In the context of machine learning, hyperparameters are indeed best defined as parameters that are external to a machine learning model. They are not learned from the training data itself but are set prior to the training process. Hyperparameters influence the training process and the performance of the model, including aspects such as the learning rate, the number of trees in a random forest, or the number of epochs for training.

They play a crucial role in model selection and optimization, as they can significantly impact how well a model performs on unseen data. By tuning hyperparameters, practitioners can improve the model’s ability to generalize beyond the training dataset.

The other options do not accurately capture the essence of hyperparameters. For instance, the notion of parameters specific to model training suggests that they are intrinsic to the training process, which is misleading since hyperparameters dictate how the training occurs rather than being parameters learned during the training itself. Similarly, parameters that remain constant throughout training could refer to fixed values that are part of the model's architecture or parameters derived from the data, but these definitions do not reflect the reality of how hyperparameters function within the machine learning pipeline.