TOPICS
Search

Vapnik-Chervonenkis Dimension


In machine learning theory, the Vapnik-Chervonenkis dimension or VC-dimension of a concept class C is the cardinality of the largest set S which can be shattered by C. If arbitrarily large sets can be shattered by C, then the VC-dimension is said to be +infty. Given a concept class C, the VC-dimension of C is sometimes denoted VCdim(C).

There are several models used to visualize the process of shattering, and hence there are a number of different models of the Vapnik-Chervonenkis dimension. In particular, it is common to use intervals (and unions thereof) on the real line, rectangles and squares in the plane, hyperplanes, etc.

There are also several results quantifying the VC-dimension to various degrees. For example, one can show that the VC-dimension of a finite concept class C satisfies

 VCdim(C)<=log_2(C).

See also

Concept, Concept Class, Order, Shattered Set

This entry contributed by Christopher Stover

Explore with Wolfram|Alpha

References

Bhaskar, A. and Sukhar, I. "VC-Dimension." 2008. http://www.cs.cornell.edu/courses/cs683/2008sp/lecture%20notes/683notes_0428.pdf.Shashua, A. "Lecture 11: PAC II." 2009. http://www.cs.huji.ac.il/~shashua/papers/class11-PAC2.pdf.

Cite this as:

Stover, Christopher. "Vapnik-Chervonenkis Dimension." From MathWorld--A Wolfram Web Resource, created by Eric W. Weisstein. https://mathworld.wolfram.com/Vapnik-ChervonenkisDimension.html

Subject classifications