What Is a Hyperparameter? — The Settings You Choose Before AI Starts Learning

When we covered machine learning earlier in this series, we talked about how a model learns by adjusting its internal connections — called parameters — based on the examples it sees during training. The model starts with random parameters and gradually improves them until it gets the right answers.

But there is another set of settings that are equally important — ones that are not learned by the model, but set by the human before training begins. These are called hyperparameters.

What is a hyperparameter?

A hyperparameter is a configuration setting that controls how the training process itself works. Not what the model learns — how it learns. The model adjusts its parameters during training. Hyperparameters are set before training starts and stay fixed throughout.

A simple analogy

Think of training a student for an exam. The student’s knowledge — what they actually learn and remember — is like the parameters. But before the studying begins, you make decisions: how many hours per day will they study? How many practice tests will they take? How long will they spend on each topic? These are the hyperparameters — they control the learning process, not the learning itself.

Common hyperparameters and what they do

Learning rate — how big a step the model takes each time it adjusts its parameters. Too large and it overshoots the right answer. Too small and it takes forever to get there. Getting this right is one of the most important decisions in training. Number of epochs — how many times the model goes through the entire training dataset. Too few and it has not learned enough. Too many and it starts memorising rather than generalising — overfitting. Batch size — how many examples the model looks at before adjusting its parameters. Larger batches are faster but may miss fine details. Smaller batches are slower but often produce better results. Number of layers — how deep the neural network is. More layers can learn more complex patterns but require more data and computing power.

Why hyperparameters matter

The same model architecture trained on the same data with different hyperparameters can produce dramatically different results. A model that performs poorly might simply have the wrong learning rate or too many epochs. Tuning hyperparameters — finding the right combination — is one of the most important and time-consuming parts of building a good AI system.

The simple rule

Parameters: what the model learns during training. Hyperparameters: the settings you choose before training that control how the learning happens. The model cannot tune its own hyperparameters — that is the human’s job.

🎁 Try MarineRef free for 1 year — use code FREE2026 at maritime.rangalabs.cloud

👉 Get one AI tip every day on WhatsApp — free. Join here: https://chat.whatsapp.com/DhaGgTuQ9GE67ykGVMgxXb