Home Glossary Item Hyperparameters
« Back to Glossary Index

Hyperparameters in the context of machine learning and artificial intelligence (AI) refer to the parameters or settings that define the structure and control the behavior of learning algorithms. Unlike other parameters in machine learning models, hyperparameters are not learned from data but are set prior to the commencement of the learning process. Hyperparameters are the “knobs” that one must adjust when setting up the machine learning algorithm. Examples of hyperparameters include the learning rate for models like linear regression, the depth of trees in decision trees, or the number of hidden layers in neural networks.


Hyperparameters play a crucial role in the modeling process as they can greatly affect the learning capability of the model. For instance, in neural networks, a hyperparameter such as the number of hidden layers can dictate the complexity of patterns that the network can learn. Similarly, in a decision tree model, the maximum depth of the tree, a key hyperparameter, can control how finely the model splits the data to make a decision.

Given their importance, selecting the right hyperparameters for a model is a critical step. This is often achieved through hyperparameter tuning or optimization techniques. This involves testing different combinations of hyperparameters on a validation set to find the mix that delivers the optimal model performance. The goal is to strike the right balance between underfitting, where the model is too simplistic, and overfitting, where it over-learns from the training data and performs poorly on new data. Hyperparameters are foundational to the proper functioning and performance of machine learning models.

« Back to Glossary Index