Table of Contents
Fetching ...

Schur--Weyl duality for diagonalizing a Markov chain on the hypercube

Persi Diaconis, Andrew Lin, Arun Ram

TL;DR

The paper develops an explicit, orthonormal eigenbasis for the Burnside Markov chain on the binary hypercube by leveraging Schur–Weyl duality and an $\mathfrak{sl}_2$-action to decompose $V^{\otimes n}$ into isotypic components labeled by Young tableaux. It introduces the $g_Q^{m,\ell}$ and $f_Q^{m,\ell}$ vectors that span irreducible subspaces and are eigenvectors of the transition operator, with eigenvalues determined by $m+\ell$ and independent of $K_n$. The resulting framework yields exact norm formulas, links to Hahn and Chebyshev polynomials, and sharp mixing-time bounds, including a bounded-time convergence from the one-one state $e_n$ in $\ell^2$ and $\ell^1$. The approach also clarifies eigenvalue multiplicities across irreducible components and opens avenues for generalizations to other $(C_k^n, S_n)$ Burnside processes. Overall, the work provides a principled, algebraic route to precise convergence analyses in highly symmetric Markov chains.

Abstract

We show how the tools of modern algebraic combinatorics -- representation theory, Murphy elements, and particularly Schur--Weyl duality -- can be used to give an explicit orthonormal basis of eigenfunctions for a "curiously slowly mixing Markov chain" on the space of binary $n$-tuples. The basis is used to give sharp rates of convergence to stationarity.

Schur--Weyl duality for diagonalizing a Markov chain on the hypercube

TL;DR

The paper develops an explicit, orthonormal eigenbasis for the Burnside Markov chain on the binary hypercube by leveraging Schur–Weyl duality and an -action to decompose into isotypic components labeled by Young tableaux. It introduces the and vectors that span irreducible subspaces and are eigenvectors of the transition operator, with eigenvalues determined by and independent of . The resulting framework yields exact norm formulas, links to Hahn and Chebyshev polynomials, and sharp mixing-time bounds, including a bounded-time convergence from the one-one state in and . The approach also clarifies eigenvalue multiplicities across irreducible components and opens avenues for generalizations to other Burnside processes. Overall, the work provides a principled, algebraic route to precise convergence analyses in highly symmetric Markov chains.

Abstract

We show how the tools of modern algebraic combinatorics -- representation theory, Murphy elements, and particularly Schur--Weyl duality -- can be used to give an explicit orthonormal basis of eigenfunctions for a "curiously slowly mixing Markov chain" on the space of binary -tuples. The basis is used to give sharp rates of convergence to stationarity.
Paper Structure (7 sections, 11 theorems, 133 equations, 2 figures)

This paper contains 7 sections, 11 theorems, 133 equations, 2 figures.

Key Result

Theorem 1.1

For the binary Burnside process started from the state $e_n = (0, 0, \cdots, 1)$ (or any other state with a single $1$), a constant number of steps is necessary and sufficient in $\ell^2$ (and therefore also in $\ell^1$). More precisely, for all $n, s \ge 3$, the chi-square distance to stationarity

Figures (2)

  • Figure 1: The eight vectors $g_Q^{m, \ell}$ for $n = 3$. Observe that we have $g_Q^{m, \ell} \in V^{(m+\ell)}$ in all cases; that is, the vector $g_Q^{m, \ell}$ is only supported on the states $S$ where $|S| = m+\ell$.
  • Figure 2: The eight vectors $f_Q^{m, \ell}$ for $n = 3$. Observe that the vector $f_Q^{m, \ell}$ with $m + \ell = 0$ is an eigenvector of $K_3$ with eigenvalue $\beta_0 = 1$, the three vectors with $m + \ell = 2$ are eigenvectors with eigenvalue $\beta_1 = \frac{1}{4}$, and all other vectors are eigenvectors with eigenvalue $0$.

Theorems & Definitions (27)

  • Theorem 1.1
  • Lemma 2.1
  • proof
  • Theorem 2.2
  • Remark
  • Proposition 2.3: diaconislinram1
  • Remark
  • Lemma 3.1
  • proof
  • proof : Proof of \ref{['orthoevthm']}
  • ...and 17 more