A full complexity dichotomy for immanant families

TL;DR

This paper establishes a comprehensive complexity classification for immanants based on the growth of a parameter related to their partitions, showing that large growth leads to computational hardness under standard complexity assumptions.

Contribution

It provides a full dichotomy for the complexity of immanants, extending previous partial results by linking growth of partition parameters to computational hardness.

Findings

Polynomial-time computability for bounded parameter families.

Hardness results for unbounded parameter families under FPT ≠ #W[1].

Hardness for polynomially growing parameters under #P and VNP.

Abstract

Given an integer $n\geq 1$ and an irreducible character $\chi_{\lambda}$ of $S_{n}$ for some partition $\lambda$ of $n$, the immanant $\mathrm{imm}_{\lambda}:\mathbb{C}^{n\times n}\to\mathbb{C}$ maps matrices $A\in\mathbb{C}^{n\times n}$ to $\mathrm{imm}_{\lambda}(A)=\sum_{\pi\in S_{n}}\chi_{\lambda}(\pi)\prod_{i=1}^{n}A_{i,\pi(i)}$. Important special cases include the determinant and permanent, which are the immanants associated with the sign and trivial character, respectively. It is known that immanants can be evaluated in polynomial time for characters that are close to the sign character: Given a partition $\lambda$ of $n$ with $s$ parts, let $b(\lambda):=n-s$ count the boxes to the right of the first column in the Young diagram of $\lambda$. For a family of partitions $\Lambda$, let $b(\Lambda):=\max_{\lambda\in\Lambda}b(\lambda)$ and write Imm$(\Lambda)$ for the problem of evaluating $\mathrm{imm}_{\lambda}(A)$ on input $A$ and $\lambda\in\Lambda$. If $b(\Lambda)<\infty$, then Imm$(\Lambda)$ is known to be polynomial-time computable. This subsumes the case of the determinant. On the other hand, if $b(\Lambda)=\infty$, then previously known hardness results suggest that Imm$(\Lambda)$ cannot be solved in polynomial time. However, these results only address certain restricted classes of families $\Lambda$. In this paper, we show that the parameterized complexity assumption FPT $\neq$ #W[1] rules out polynomial-time algorithms for Imm$(\Lambda)$ for any computationally reasonable family of partitions $\Lambda$ with $b(\Lambda)=\infty$. We give an analogous result in algebraic complexity under the assumption VFPT $\neq$ VW[1]. Furthermore, if $b(\lambda)$ even grows polynomially in $\Lambda$, we show that Imm$(\Lambda)$ is hard for #P and VNP. This concludes a series of partial results on the complexity of immanants obtained over the last 35 years.