AI Glossary Term
What does “Weight” mean?
Definition
A learned numeric value that scales signals passing through a neural network.
How Weight works
Example
In a spam model, the weight on the word 'free' might be large and positive, pushing the score toward spam, while the weight on a coworker's name may be negative. Training sets these numbers from examples.
Why it matters
Weights hold what the model has learned. Releasing a model's weights means sharing the trained model.
Common questions
Weight vs parameter?
Weights are the most common kind of parameter. Biases are another.
Are weights readable?
They are just numbers, and individual ones rarely have an obvious meaning.
What are open weights?
Models whose trained weights are published so others can run and adapt them.