Hyperparameters are the settings in a machine learning model that are decided before training starts, and they help guide the learning process.
Think about when you’re baking cookies. The recipe you follow might tell you to bake them at 350 degrees for 12 minutes. But what if you like your cookies a little crunchier? You might decide to bake them at a higher temperature or for a bit longer. The temperature and time are like hyperparameters – they’re not something you learn from the recipe, but something you decide on your own based on what you want the outcome to be.
In a machine learning model, hyperparameters might include things like the learning rate, which determines how much the model changes in response to the error it made on its prediction. Or in a neural network, the number of layers in the network, and the number of nodes in each layer, are hyperparameters.
Hyperparameters are important because they can greatly affect the performance of the model. They’re often chosen through trial and error, by trying out different values and seeing which ones lead to the best performance on a validation set. This process is known as hyperparameter tuning.
« Back to Glossary Index