What Is a Hyperparameter?

A hyperparameter is a setting chosen before training that controls how an AI model learns, such as the learning rate or the number of training passes. Unlike the parameters a model learns from data, hyperparameters are set by the people building it and shape how the training process behaves.

How a hyperparameter works

A hyperparameter is configured before training begins and stays fixed while the model learns. Common examples include the learning rate, which controls how big an adjustment the model makes at each step, the batch size, which sets how many examples it looks at before updating, and the number of epochs, or passes over the data. These settings do not come from the data; they govern the learning process itself.

Getting them right matters, because poor choices lead to a model that learns too slowly, misses patterns, or overfits. Because the best values are rarely obvious, teams find them through hyperparameter tuning, testing different combinations and keeping the ones that produce the best results.

Why hyperparameters matter for AI

Hyperparameters strongly influence how well a model turns out, even though they are not learned from data. The same architecture and dataset can produce very different results depending on how the training is configured, which is why tuning them is a standard part of building a model. Good hyperparameters help a model learn the real patterns in its data without memorizing noise or missing signal.