Concept Glossary: W¶
This page lists English glossary entries for this letter. Entry bodies are generated from per-term source files with includes.
weight¶
- Meaning: A weight is a parameter that controls how strongly an input value or intermediate value affects a model's output calculation. It is a numerical handle for connection strength or influence.
- Why it matters: In many machine-learning and deep-learning models, training can be read as adjusting weights so the input-output relationship fits the task better. The concept makes "the model learns" concrete as changing numerical parameters in the calculation.
- Related concepts:
parameter,representation,model training - Core Section:
P1-4.3 - Appears in:
P1-5.1,P1-12.1,P2-11.1,P6-21.1
weighted sum¶
- Meaning: A weighted sum multiplies each input by its own weight and then adds the results into one value. For example, \(x_1w_1 + x_2w_2\) folds two inputs into one score while reflecting them with different strengths.
- Why it matters: Many AI model calculations, including linear layers, matrix multiplication, and attention scores, repeat this pattern. The term keeps the role of the weight clear: a weight is not just a number being multiplied, but a value that controls how strongly an input is reflected.
- Related concepts:
weight,linear combination,matrix multiplication,activation function - Core Section:
P2-3.3 - Appears in:
P5-1.1,P5-1.2