What theoretical measure quantifies the capacity (complexity) of a statistical classification algorithm, defined as the size of the largest set of points it can shatter?
Show answer
VC dimension
The Vapnik-Chervonenkis (VC) dimension is a core concept in statistical learning theory. It provides a way to bound the generalization error of a model based on its capacity and the amount of training data available.
Also asked as:
- What statistical learning concept is defined as the maximum number of points that a classification hypothesis class can shatter?