Methods and apparatuses for measuring diversity in combinatorial structures
View Patent ↗A method for computing a diversity measure H(m) for combinatorial structures involves identifying all M possible substructures having m elements from among the n elements of the combinatorial structure. The number of the substructures that are similar to each such substructure is determined, and the frequency of each distinct substructure is calculated using the number of similar substructures and the total number of substructures M. The method uses the frequency of each distinct substructure to compute an entropy corresponding to m. By the same process described above, and entropy corresponding to m+1 is computed. The entropy corresponding to m+1 is subtracted from the entropy corresponding to m to produce the diversity measure H(m). In the preferred embodiment, similar substructures are determined by being identical or isomorphic. In an alternative embodiment, a distance function is used to compute a distance between two substructures, and only if the distance is less than a predetermined threshold are the two substructures determined to be similar. In the preferred embodiment, the entropy is computed by summing the frequency of each distinct substructure multiplied by the logarithm of the frequency of each distinct substructure. In an alternative embodiment, the entropy is computed by summing the frequency of each distinct substructure by the logarithm of the quotient of the frequency divided by an expected frequency of the distinct substructure. Generalized graphs such as can be used to model the Web are combinatorial structures suitable for use with the methods according to the present invention.
1. A method for computing a diversity measure for a group of web pages, C, having n elements, wherein the diversity measure indicates a level of diversity for the group of web pages, the method comprising steps of:
(a) identifying M substructures c 1 through C M each having m elements from among the n elements of the group of web pages C, where M equals n!/[(n−m)!m!];
(b) for each substructure c i , for i from l to M, determining a number n i of the M substructures c i through c M that are similar to the substructure c i ; and
(c) computing a first entropy Φ(m) based upon all the numbers n i and M computed during step (d) and based upon M in computed step (a).
2. A method as in claim 1 , further comprising the steps of:
(d) repeating steps (a) and (b) with m+1 substituted for m;
(e) computing a second entropy φ(m+1) based upon all the numbers n i and M computed during step (d); and
(f) subtracting the second entropy φ(m+1) from the first entropy φ(m) to produce the diversity measure.
3. A method as in claim 2 , wherein steps (c) and (e) comprise the steps of:
for each i from 1 to M:
computing a fraction F i by dividing n i by M; and
computing a logarithm of fraction F i ;
computing a sum by adding all logarithms of fractions F i for i from l to M; ad dividing the sum by M.
4. A method as in claim 2 , wherein step (b) comprises the steps of, for each substructure c i for i from l to M:
for each substructure c j for j from l to M:
computing a distance function d(c i ,c j ) representing a measure of a difference between substructure c i and substructure c j ;
comparing the distance function d(c i ,c j ) to a threshold; and
determining the substructures c i and c j to be similar if and only if the distance function d(c i ,c j ) is less than the threshold.
5. A method as in claim 2 , wherein steps (c) and (e) comprise the steps of:
for each distinct substructure c i ;
computing a frequency f i by dividing n i by M;
computing a logarithm of frequency f i ; and
computing a product by multiplying the frequency f i and the logarithm of frequency f i ; and
computing a sum by adding all products of the frequencies f i and the logarithms of frequencies f i .
6. A method as in claim 2 , wherein step (b) comprises the steps of:
for each substructure c i for i from l to M:
monotonically renumbering m elements of c i from l to m; and
for each substructure c j for j from l to M:
monotonically renumbering m elements of c j from l to m; and
determining the substructures c i and c j to be similar if and only if they are identical.
7. A method as in claim 2 , wherein step (b) comprises the steps of:
for each substructure c i for i from l to M:
monotonically renumbering m elements of c i from l to m; and
for each substructure c j for j from l to M:
monotonically renumbering m elements of c i from l to m; and
determining the substructures c i and c j to be similar if and only if they are identical or isomorphic.
8. A method as in claim 2 , wherein steps (c) and (c) comprise the steps of:
for each distinct substructure c i :
computing a frequency f i by dividing n i by M;
computing a quotient by dividing the frequency f i by an expected frequency p i ;
computing a logarithm of quotient q i ; and
computing a product by multiplying the frequency f i and the logarithm of quotient q i ; and
computing a sum by adding all products of the frequencies f i and the logarithms of quotients q i .
9. A computer readable storage medium, comprising:
computer readable program code embodied on said computer readable storage medium, said computer readable program code for programming a computer to perform a method for computing a diversity measure for a group of web pages, C, having n elements, wherein the diversity measure indicates a level of diversity for the group of web pages, the method comprising steps of:
(a) identifying M substructures c l through c M each having m elements from among the n elements of the group of web pages C, where M equals n!/[(n−m)!m!];
(b) for each substructure c i , for i from l to M, determining a number n i of the M substructures c l through c M that are similar to the substructure c i ; and
(c) computing a first entropy Φ(m) based upon all the numbers n i computed during step (b) and based upon M in computed step (a).
10. A computer readable storage medium as in claim 9 , the method further comprising the steps of:
(d) repeating steps (a) and (b) with m+1 substituted for m;
(e) computing a second entropy Φ(m+1) based upon all the numbers n i and M computed during step (d); and
subtracting the second entropy Φ(m+1) from the fist entropy Φ(m) to produce the diversity measure.
11. A computer readable storage medium as in claim 10 , wherein steps (c) and (e) comprise the steps of:
for each i from l to M:
computing a fraction P i by dividing n j by M; and
computing a logarithm of fraction F i ;
computing a sum by adding all logarithms of fractions F i for i from l to M; and
dividing the sum by M.
12. A computer readable storage medium as in claim 10 , wherein step (b) comprises the steps of, for each substructure c i for i from l to M:
for each substructure c j for j from l to M:
computing a distance function d(c i ,c j ) representing a measure of a difference between substructure c i and substructure c j ;
comparing the distance function d(c i ,c j ) to a threshold; and
determining the substructures c i and c j to be similar if and only if the distance function d(c i ,c j ) is less than the threshold.
13. A computer readable storage medium as in claim 10 , wherein steps (c) and (e) comprise the steps of:
for each distinct substructure c i :
computing a frequency f i by dividing n i by M;
computing a logarithm of frequency f i ; and
computing a product by multiplying the frequency f i and the logarithm of frequency f i ; and
computing a sum by adding all products of the frequencies f i and the logarithms of frequencies f j .
14. A computer readable storage medium as in claim 10 , wherein step (b) comprises the steps of:
for each substructure c i for i from l to M:
monotonically renumbering m elements of c i from l to m; and
for each substructure c j for j from l to M:
monotonically renumbering m elements of c j from l to m; and
determining the substructures c i and c j to be similar if and only if they are identical.
15. A computer readable storage medium as in claim 10 , wherein step (b) comprises the steps of:
for each substructure c i for i from l to M:
monotonically renumbering m elements of c i from l to m; and
for each substructure c j for j from l to M:
monotonically renumbering m elements of c j from l to m; and
determining the substructures c i and c j to be similar if and only if they are identical or isomorphic.
16. A computer readable storage medium as in claim 10 , wherein steps (c) and (e) comprise the steps of:
for each distinct substructure c i :
computing a frequency f i by dividing n i by M;
computing a quotient by dividing the frequency f i by an expected frequency p i ;
computing a logarithm of quotient q i ; and
computing a product by multiplying the frequency f i and the logarithm of quotient q i ; and
computing a sum by adding all products of the frequencies f i and the logarithms of quotients q i .
17. A computer system, comprising:
a processor; and
a processor readable storage medium coupled to the processor having processor readable program code embodied on said processor readable storage medium, said processor readable program code for programming the computer system to perform a method for computing a diversity measure for a group of web pages, C, having n elements, wherein the diversity measure indicates a level of diversity for the group of web pages, the method comprising steps of:
(a) identifying M substructures c l , through c M each having m elements from among the n elements of the group of web pages C, where M equals n!/[(n−m)!m!];
(b) for each substructure c i , for i from l to M, determining a number n i of the M substructures c l through c M that are similar to the substructure c i ; and
(c) computing a first entropy Φ(m) based upon all the numbers n i computed during step (b) and based upon M in computed step (a).
18. A computer system as in claim 17 , the method further comprising the steps of:
(d) repeating steps (a) and (b) with m+1 substituted for m;
(e) computing a second entropy Φ(m+1) based upon all the numbers n i and M computed during step (d); and
(f) subtracting the second entropy Φ(m+1) from the first entropy Φ(m) to produce the diversity measure.
19. A computer system as in claim 18 , wherein steps (c) and (e) comprise the steps of:
for each i from l to M:
computing a fraction F i by dividing n i by M; and
computing a logarithm of fraction F i ;
computing a sum by adding all logarithms of fractions F i for i from l to M; and
dividing the stun by M.
20. A computer system as in claim 18 , wherein step (b) comprises the steps of, for each substructure c i for i from l to M:
for each substructure C j for j from l to M:
computing a distance function d(c i ,c j ) representing a measure of a difference between substructure c l and substructure c j ;
comparing the distance function d(c i ,c j ) to a threshold; and
determining the substructures c i and c j to be similar if and only if the distance function d(c i ,c j ) is less than the threshold.
21. A computer system as in claim 18 , wherein steps (c) and (e) comprise the steps of:
for each distinct substructure c i :
computing a frequency f i by dividing n i by M;
computing a logarithm of frequency f i ; and
computing a product by multiplying the frequency f i and the logarithm of frequency f i ; and
computing a sum by adding all products of the frequencies f i and the logarithms of frequencies f i .
22. A computer system as in claim 18 , wherein step (b) comprises the steps of:
for each substructure c i for i from l to M:
monotonically renumbering m elements of c i from l to m; and
for each substructure c j for j from l to M:
monotonically renumbering m elements of c j from l to m; and
determining the substructures c i and c j to be similar if and only if they are identical.
23. A computer system as in claim 18 , wherein step (b) comprises the steps of:
for each substructure c i for i from l to M:
monotonically renumbering m elements of c j from l to m; and
for each substructure c j for j from l to M:
monotonically renumbering m elements of c j from l to m; and
determining the substructures c i and c j to be similar if and only if they are identical or isomorphic.
24. A computer system as in claim 18 , wherein steps (c) and (e) comprise the steps of:
for each distinct substructure c i :
computing a frequency f i by dividing n i by M;
computing a quotient by dividing the frequency f i by an expected frequency p i ;
computing a logarithm of quotient q i ; and
computing a product by multiplying the frequency f i and the logarithm of quotient q i ; and
computing a sum by adding all products of the frequencies f i and the logarithms of quotients q i .