Regression and VC dimension

[Mikhail]

In this course VC dimension was introduced by means of maximal number of points can be shuttered by the hypothesis set (if I'm not mistaken). It is clear for the classification, that we deal with dichotomies of {+1, -1} points. Well, but how does it relate to the regression model, when we learn f: R^n -> R? Clarify, please.

Thanks.

[yaser]

Originally Posted by Mikhail

In this course VC dimension was introduced by means of maximal number of points can be shuttered by the hypothesis set (if I'm not mistaken). It is clear for the classification, that we deal with dichotomies of {+1, -1} points. Well, but how does it relate to the regression model, when we learn f: R^n -> R? Clarify, please.

Thanks.

The definition is technically modified, borrowing some concepts from measure theory. Look up the book by Vapnik.