Discussion:
combinatorial interpretation of the cycleIndex series
Martin Rubey
2007-06-06 09:12:45 UTC
Permalink
Dear all,

Browsing Marni Mishna's thesis, I stumbled over the following very very nice
observation:

Regard the cycleIndexSeries as series in the powersum symmetric functions.
I.e., write p_i instead of x_i, as I suggested many times already.

The p_i are functions in variables t1, t2, t3, ... Expand the cycleIndexSeries
in these variables.

Then the coefficient of

t^n = t1^n1 t2^n2 t3^n3 ...

is the number of non-isomorphic colored structures using n1 times color 1, n2
times color 2, n3 times color 3, ...

-------------------------------------------------------------------------------
For example, take the cycleIndexSeries of E_2, or, in AldorCombinat parlance,
RestrictedSpecies(SetSpecies, 2):

(12) -> E2 ==> Interpret([parse "RestrictedSpecies(SetSpecies, 2)"], ACINT)
Type: Void
(13) -> coefficient(cycleIndexSeries()$E2,2)

1 2 1
(13) - x + - x
2 1 2 2
Type: SparseDistributedPolynomial(ACFraction ACInteger,CycleIndexVariable,SparseIndexedPowerProduct(CycleIndexVariable,ACMachineInteger))


Interpreting x1 as t1+t2+t3+... and x2 as t1^2+t2^2+t3^2+... we obtain

t1^2 + t2^2 + t3^2 + ... + t1 t2 + t1 t3 + t2 t3 + ...

Now, the combinatorial interpretation is as follows: there is exactly one E2
structure on {a,b}, namely {{a,b}}. We can color this (using the positive
integers as colors) as follows:

{1, 1}, {2, 2}, {3, 3}, ... {1, 2}, {1, 3}, {2, 3}, ...

By the way, if we had a package for symmetric functions, we could transform
1/2(p1^2 + p2) into h2, i.e., the complete symmetric function, from where the
combinatorial interpretation is obvious.

-------------------------------------------------------------------------------
For another example, take the cycleIndexSeries of C_3, cycles on three
elements 1/3 p_1^3 + 2/3 p_3. Expanding in terms of t1, t2, t3,... we obtain

t1^3 + t2^3 + t3^3 + ...

+ t1^2 t2 + t1 t2^2 + ...

+ 2 t1 t2 t3 + 2 t1 t2 t4 + ...

There are two cycles on {a,b,c}, namely (abc) and (acb). Coloring with one or
two colors, there is only one such colored structure:


1

/ \

1 --- 2

But when we color with three colors, we get two different structures:

1 1

/ \ / \

3 --- 2 2 --- 3

By the way, if we had a package for symmetric functions, we could transform 1/3
p_1^3 + 2/3 p_3 into s_{1,1,1} + s_{3}, i.e., the Schur functions, from where
the combinatorial interpretation is again obvious: semistandard Young tableaux
like

1 or 1 1 1 or 1 2 2 or 1 1 2 or 1 2 3.
2
3

(t1 t2 t3 t1^3 t1 t2^2 t1^2 t2 t1 t2 t3)

-------------------------------------------------------------------------------

Marni Mishna writes that this is due to Cauchy-Frobenius, i.e., (I guess)
Proposition 4.3.2 in BLL, expressing the cycle index series as sum over all
isomorphism types of the cycle index polynomials. I still have to check that.

In any case, this should be convincing evidence to write p_i instead of x_i and
interpret the cycle index series as a symmetric function.

Martin


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
Ralf Hemmecke
2007-06-06 09:38:53 UTC
Permalink
Post by Martin Rubey
In any case, this should be convincing evidence to write p_i instead of x_i and
interpret the cycle index series as a symmetric function.
You are probably not saying that instead of x you want to use the letter
p instead. That would be ridiculous.

As you have shown, one only has to take the CIS, substitute for the x_i
some symmetric polynomials and voila, you get what you want. Why is that
so problematic?

As I understand BLL, we should rather consider (extend AC to) weighted
species. That is how you get your substitutions. Am I wrong?

Ralf

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
Martin Rubey
2007-06-06 09:44:21 UTC
Permalink
Post by Martin Rubey
In any case, this should be convincing evidence to write p_i instead of x_i
and interpret the cycle index series as a symmetric function.
You are probably not saying that instead of x you want to use the letter p
instead. That would be ridiculous.
Well, perhaps it sounds ridiculous, but I am saying exactly that. And,
additionally, we should keep in mind that we really are working with symmetric
functions.
As you have shown, one only has to take the CIS, substitute for the x_i some
symmetric polynomials and voila, you get what you want. Why is that so
problematic?
It isn't. Only, good notation is often very useful. You see, substituting for
the x_i other symmetric functions (not polynomials, by the way), we do not get
anything sensible.
As I understand BLL, we should rather consider (extend AC to) weighted
species. That is how you get your substitutions. Am I wrong?
No, I don't think so, but to be honest, I don't know. In any case, we cannot
start yet another project. multisort is difficult enough, and in any case far
more important than weights.

Martin


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
Ralf Hemmecke
2007-06-06 09:54:51 UTC
Permalink
Post by Martin Rubey
Post by Ralf Hemmecke
As I understand BLL, we should rather consider (extend AC to) weighted
species. That is how you get your substitutions. Am I wrong?
No, I don't think so, but to be honest, I don't know. In any case, we cannot
start yet another project. multisort is difficult enough, and in any case far
more important than weights.
I just wanted to say, that instead of restricting to symmetric
functions, I would rather implement the concept of weighted species.
Later (than multisort), of course, but I think, it is useful, too.

Ralf

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
Martin Rubey
2007-06-06 11:21:08 UTC
Permalink
Post by Martin Rubey
Marni Mishna writes that this is due to Cauchy-Frobenius, i.e., (I guess)
Proposition 4.3.2 in BLL, expressing the cycle index series as sum over all
isomorphism types of the cycle index polynomials. I still have to check that.
OK, I checked. It is indeed due to 4.3.2 and Theorem 17 of the Appendix in
BLL, i.e. Polya-Redfield. If you prefer ECII by Stanley, it is Theorem 7.24.4.

I'll put that into the documentation as soon as I can.

Martin


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/

Loading...