The growth function, also called the shatter coefficient or the shattering number, measures the richness of a set family or class of function. It is especially used in the context of statistical learning theory, where it is used to study properties of statistical learning methods. The term 'growth function' was coined by Vapnik and Chervonenkis in their 1968 paper, where they also proved many of its properties.[1] It is a basic concept in machine learning.[2] [3]

Definitions

edit

Set-family definition

edit

Let   be a set family (a set of sets) and   a set. Their intersection is defined as the following set-family:

 

The intersection-size (also called the index) of   with respect to   is  . If a set   has   elements then the index is at most  . If the index is exactly 2m then the set   is said to be shattered by  , because   contains all the subsets of  , i.e.:

 

The growth function measures the size of   as a function of  . Formally:

 

Hypothesis-class definition

edit

Equivalently, let   be a hypothesis-class (a set of binary functions) and   a set with   elements. The restriction of   to   is the set of binary functions on   that can be derived from  :[3]: 45 

 

The growth function measures the size of   as a function of  :[3]: 49 

 

Examples

edit

1. The domain is the real line  . The set-family   contains all the half-lines (rays) from a given number to positive infinity, i.e., all sets of the form   for some  . For any set   of   real numbers, the intersection   contains   sets: the empty set, the set containing the largest element of  , the set containing the two largest elements of  , and so on. Therefore:  .[1]: Ex.1  The same is true whether   contains open half-lines, closed half-lines, or both.

2. The domain is the segment  . The set-family   contains all the open sets. For any finite set   of   real numbers, the intersection   contains all possible subsets of  . There are   such subsets, so  . [1]: Ex.2 

3. The domain is the Euclidean space  . The set-family   contains all the half-spaces of the form:  , where   is a fixed vector. Then  , where Comp is the number of components in a partitioning of an n-dimensional space by m hyperplanes.[1]: Ex.3 

4. The domain is the real line  . The set-family   contains all the real intervals, i.e., all sets of the form   for some  . For any set   of   real numbers, the intersection   contains all runs of between 0 and   consecutive elements of  . The number of such runs is  , so  .

Polynomial or exponential

edit

The main property that makes the growth function interesting is that it can be either polynomial or exponential - nothing in-between.

The following is a property of the intersection-size:[1]: Lem.1 

  • If, for some set   of size  , and for some number  ,   -
  • then, there exists a subset   of size   such that  .

This implies the following property of the Growth function.[1]: Th.1  For every family   there are two cases:

  • The exponential case:   identically.
  • The polynomial case:   is majorized by  , where   is the smallest integer for which  .

Other properties

edit

Trivial upper bound

edit

For any finite  :

 

since for every  , the number of elements in   is at most  . Therefore, the growth function is mainly interesting when   is infinite.

Exponential upper bound

edit

For any nonempty  :

 

I.e, the growth function has an exponential upper-bound.

We say that a set-family   shatters a set   if their intersection contains all possible subsets of  , i.e.  . If   shatters   of size  , then  , which is the upper bound.

Cartesian intersection

edit

Define the Cartesian intersection of two set-families as:

 .

Then:[2]: 57 

 

Union

edit

For every two set-families:[2]: 58 

 

VC dimension

edit

The VC dimension of   is defined according to these two cases:

  • In the polynomial case,   = the largest integer   for which  .
  • In the exponential case  .

So   if-and-only-if  .

The growth function can be regarded as a refinement of the concept of VC dimension. The VC dimension only tells us whether   is equal to or smaller than  , while the growth function tells us exactly how   changes as a function of  .

Another connection between the growth function and the VC dimension is given by the Sauer–Shelah lemma:[3]: 49 

If  , then:
for all  :  

In particular,

for all  :  
so when the VC dimension is finite, the growth function grows polynomially with  .

This upper bound is tight, i.e., for all   there exists   with VC dimension   such that:[2]: 56 

 

Entropy

edit

While the growth-function is related to the maximum intersection-size, the entropy is related to the average intersection size:[1]: 272–273 

 

The intersection-size has the following property. For every set-family  :

 

Hence:

 

Moreover, the sequence   converges to a constant   when  .

Moreover, the random-variable   is concentrated near  .

Applications in probability theory

edit

Let   be a set on which a probability measure   is defined. Let   be family of subsets of   (= a family of events).

Suppose we choose a set   that contains   elements of  , where each element is chosen at random according to the probability measure  , independently of the others (i.e., with replacements). For each event  , we compare the following two quantities:

  • Its relative frequency in  , i.e.,  ;
  • Its probability  .

We are interested in the difference,  . This difference satisfies the following upper bound:

 

which is equivalent to:[1]: Th.2 

 

In words: the probability that for all events in  , the relative-frequency is near the probability, is lower-bounded by an expression that depends on the growth-function of  .

A corollary of this is that, if the growth function is polynomial in   (i.e., there exists some   such that  ), then the above probability approaches 1 as  . I.e, the family   enjoys uniform convergence in probability.

References

edit
  1. ^ a b c d e f g h Vapnik, V. N.; Chervonenkis, A. Ya. (1971). "On the Uniform Convergence of Relative Frequencies of Events to Their Probabilities". Theory of Probability & Its Applications. 16 (2): 264. doi:10.1137/1116025. This is an English translation, by B. Seckler, of the Russian paper: "On the Uniform Convergence of Relative Frequencies of Events to Their Probabilities". Dokl. Akad. Nauk. 181 (4): 781. 1968. The translation was reproduced as: Vapnik, V. N.; Chervonenkis, A. Ya. (2015). "On the Uniform Convergence of Relative Frequencies of Events to Their Probabilities". Measures of Complexity. p. 11. doi:10.1007/978-3-319-21852-6_3. ISBN 978-3-319-21851-9.
  2. ^ a b c d Mohri, Mehryar; Rostamizadeh, Afshin; Talwalkar, Ameet (2012). Foundations of Machine Learning. US, Massachusetts: MIT Press. ISBN 9780262018258., especially Section 3.2
  3. ^ a b c d Shalev-Shwartz, Shai; Ben-David, Shai (2014). Understanding Machine Learning – from Theory to Algorithms. Cambridge University Press. ISBN 9781107057135.