Definition
Hypothesis (Machine Learning)
A hypothesis is a single specific function that maps inputs to outputs, chosen by a learning algorithm from a model (hypothesis class).
A model defines the space of possibilities. The hypothesis is the point within that space that the algorithm selects after training on data.
Hypothesis vs Model
| Hypothesis | Model | |
|---|---|---|
| What it is | a single specific function | the set of all candidate functions |
| When it matters | after training (the result) | before training (defines capacity) |
| Example | linear regression: all lines | |
| Synonym | — | hypothesis class |