Abstract:Left-right and conjugation actions on matrix tuples have received considerable attention in theoretical computer science due to their connections with polynomial identity testing, group isomorphism, and tensor isomorphism. In this paper, we present polynomial-time algorithms for computing canonical forms of matrix tuples over a finite field under these actions. Our algorithm builds upon new structural insights for matrix tuples, which can be viewed as a generalization of Schur's lemma for irreducible representations to general representations.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: on a finite field, how to calculate the canonical forms for matrix tuples under the left - right action and the conjugation action, and ensure that the algorithm is completed within polynomial time.
### Problem Background
Matrix tuples have received extensive attention in theoretical computer science because they are closely related to problems such as polynomial identity testing, group isomorphism, and tensor isomorphism. These problems are not only of great significance in mathematics but also have wide applications in computer science.
### Specific Problem Description
1. **Canonical Forms under Left - Right Action**:
- For a matrix tuple \((A_1, A_2,\ldots, A_{\ell})\) of size \(n\times m\), the left - right action is defined by two invertible matrices \(L\in\mathbb{F}_q^{n\times n}\) and \(R\in\mathbb{F}_q^{m\times m}\), which transform \((A_1, A_2,\ldots, A_{\ell})\) into \((L A_1 R^{-1}, L A_2 R^{-1},\ldots, L A_{\ell} R^{-1})\).
- The goal is to find a polynomial - time algorithm such that for any two equivalent matrix tuples \(A\) and \(A'\), the algorithm outputs the same canonical form \(A^*\).
2. **Canonical Forms under Conjugation Action**:
- For a square matrix tuple \((A_1, A_2,\ldots, A_{\ell})\) of size \(n\times n\), the conjugation action is defined by an invertible matrix \(L\in\mathbb{F}_q^{n\times n}\), which transforms \((A_1, A_2,\ldots, A_{\ell})\) into \((L A_1 L^{-1}, L A_2 L^{-1},\ldots, L A_{\ell} L^{-1})\).
- The goal is also to find a polynomial - time algorithm such that for any two conjugate matrix tuples \(A\) and \(A'\), the algorithm outputs the same canonical form \(A^*\).
### Solution
The main contribution of the paper is to propose a new structural insight and design a polynomial - time algorithm based on it to calculate the canonical forms of matrix tuples. Specifically:
- **Structural Insight**: By generalizing Schur's lemma, the author proposes new structural results regarding matrix tuples. These results reveal the internal structure of matrix tuples under different actions and provide a theoretical basis for designing efficient canonical form algorithms.
- **Algorithm Design**: Based on the above - mentioned structural results, the author designs a randomized Las - Vegas algorithm that can calculate the canonical forms of matrix tuples within polynomial time. The core of this algorithm is to gradually construct the required row vector sub - tuples corresponding to the indecomposable blocks (IBC - tuple) by detecting the eigenspaces of the row vector subspaces, thereby achieving the calculation of the canonical form.
### Applications and Significance
This result not only solves important problems in theoretical computer science but also has potential application values in multiple fields, for example:
- **Symbolic Determinant Identity Testing (SDIT)**: Studying the equivalence classes of matrix tuples is helpful for understanding the SDIT problem.
- **Group Isomorphism Testing**: It improves the time complexity of p - group isomorphism testing.
- **Tensor Isomorphism and Canonical Forms**: It provides an important intermediate step for tensor isomorphism and canonical form problems.
In summary, this paper provides a brand - new solution for calculating the canonical forms of matrix tuples by proposing new structural results and efficient algorithms.