Path: blob/main/course-contents/2024-02-07--notes-RT--functions.md
908 views
------\newcommand{\trivial}{\mathbf{1}}
Irreducible representations of finite abelian groups
Let be a finite abelian group, written additively.
We set for the set of all group homomorphisms . We can make into a group by declaring for that is the mapping .
Proposition: : is an abelian group, and .
Sketch: : When is cyclic let be primitive th root of unity.
Observe that elements of determine 1-dimensional representations, which are necessarily irreducible. Since a 1-dimensional representation coincides with its trace, our results on the characters of irreducible representations imply that the functions -valued functions form an orthonormal basis for .
Let us write where .
Given a function , i.e. an element , we may write
The Fourier transform of is the function given by
Remark: : When , one often views the values for as samples of some periodic function of (say) a real variable , viewed as time.
Remark: : In any event, for abelian we have two natural bases for : the functions for , and the functions .
How to study for non-abelian ?
Idea: the matrix coefficients of linear representations define functions on . If is a linear representation, let be a basis of and let be the dual basis.
For we get a function by the rule
Claim: : Let be an irreducible representation and let be the isotypic component of the regular representation. Then the functions provide a basis for where defines the irreducible representation dual to .
Irreducible representations of the symmetric group
Recall that two elements are conjugate they have the same cycle structure. We state this in the following form:
Lemma: : Conjugacy classes in are in bijection with partitions .
For example, there are partitions of ; they are , , , , and .
As a consequence, we know:
Proposition: : Isomorphism classes of irreducible representations of on complex vector spaces are in bijection with partitions .
Here is a quick overview of some facts about the representations of , without proofs:
In fact, for each partition , there is a construction of an irreducible representation for . To begin the construction, associate with the subgroup let be the set of left cosets of in .
Thus is a set on which the symmetric group acts, and we may consider the permutation representation , and for brevity we write .
One defines the dominance ordering on partitions of by the rule if and only if for each
There is a labeling of irreducible representations of by partition; let us write for the irreducible representation corresponding to . It is known as a Specht representation.
Theorem: : and .
As stated, the Theorem appears to depend on our knowledge of the irreducible representations, but in fact it gives us a way to define them.
For a fixed , consider all homomorphisms of -representations $$M^\mu \to M^\lambda \quad \text{ for $\lambda \trianglelefteq \muParseError: KaTeX parse error: Expected 'EOF', got '}' at position 1: }̲;R^\lambda \subset M^\lambda$ for the sum of the image of all of these homomorphisms.
The Theorem implies that is irreducible.
Here are some special cases:
if then ParseError: KaTeX parse error: Undefined control sequence: \trivial at position 25: … = S^\lambda = \̲t̲r̲i̲v̲i̲a̲l̲ is the trivial representation.
if them is the (1-dimensional) sign representation.
if then is the permutation representation , and ParseError: KaTeX parse error: Undefined control sequence: \trivial at position 30: …\lambda \oplus \̲t̲r̲i̲v̲i̲a̲l̲ so that .
Observe that indeed .
when , we saw previously that has 3 irreducible representations. They can be described as
ParseError: KaTeX parse error: Undefined control sequence: \trivial at position 11: S^{(3)} = \̲t̲r̲i̲v̲i̲a̲l̲ the trivial representation
the sign representation
, an irreducible representation of dimension 2
Rank preferences
I want to describe in brief the ideas investigated in [@diaconisGeneralizationSpectralAnalysis1989], which amounts as an application of representation theory on the symmetric group to statistics.
First, an example
Suppose we were to ask people to rank their preferred ice cream flavors from the following ordered list:
[
pistachio,chocolate,strawberry,vanilla,neopolitan]
Numbering the flavors [ 0, 1, 2, 3, 4 ] in order, we can represent an individual's preference using a permutation .
For example the preference list
[
neopolitan,strawberry,chocolate,vanilla,pistachio]
corresponds to the product of transposition .
So our survey data amounts to a list of elements .
Ranking in more generality
In general, will denote the number of items to be ranked, and we assume given a list of ranking data:
The main statistic of interest is the frequency function given for by the rule .
Idea: View as a vector in the regular representation . We want to understand how decomposes in some natural descriptions of .
Bibliography
::: {.refs} :::