Skip to content
BENIBIX Learn

Model Parameter

Also known as Parameter (Machine Learning), Learned Parameter

A number inside a model that training learns. Together they hold what it knows.

Think of it like

Think of it like the knobs on a radio. Training turns them until the sound is clear.

Example

Jack's line for house prices has two parameters: a starting price and a price per square metre. Training finds them from his data.

Why it matters

When people say a model is big, they mean it has many parameters. More can mean more skill, and always means more cost to run.

Definition

A model parameter is one of the numbers a model learns in training. Together, they hold all the model knows. Training starts them at random. Then it nudges them, step by step, until the answers are good. A small model may have a few, and a big one has billions. Nobody sets them by hand.
The numbers inside the box are parameters. Training fills them in, and you never touch them.

Common mistake

Mixing up parameters and hyperparameters. The model learns the first kind. You pick the second kind by hand, before it starts.

Key takeaways

  1. 1
    Learned, not set. Training finds the values.
  2. 2
    Hold the knowledge. All the model knows is in them.
  3. 3
    Size measure. A big model has billions.

Compare

Learn it properly