On Recurrence Relations of Multi-dimensional Sequences

Hamid Rahkooy
2024-10-23
Abstract:In this paper, we present a new algorithm for computing the linear recurrence relations of multi-dimensional sequences. Existing algorithms for computing these relations arise in computational algebra and include constructing structured matrices and computing their kernels. The challenging problem is to reduce the size of the corresponding matrices. In this paper, we show how to convert the problem of computing recurrence relations of multi-dimensional sequences into computing the orthogonal of certain ideals as subvector spaces of the dual module of polynomials. We propose an algorithm using efficient dual module computation algorithms. We present a complexity bound for this algorithm, carry on experiments using Maple implementation, and discuss the cases when using this algorithm is much faster than the existing approaches.
Symbolic Computation,Commutative Algebra
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to calculate the linear recurrence relations of multi - dimensional sequences. Specifically, the author proposes a new algorithm for calculating the linear recurrence relations of multi - dimensional sequences, which are very important in computational algebra. ### Problem Background Existing algorithms for calculating recurrence relations usually involve constructing structured matrices and calculating their kernels. However, one challenge faced by these methods is how to reduce the size of the relevant matrices, thereby improving computational efficiency. To this end, the author proposes a new method, which transforms the problem of calculating the recurrence relations of multi - dimensional sequences into the problem of calculating the orthogonal spaces of certain ideals in the polynomial dual module. ### Main Contributions 1. **Problem Transformation**: The author shows how to transform the problem of calculating the recurrence relations of multi - dimensional sequences into the problem of calculating the orthogonal spaces of certain ideals in the dual module. 2. **Algorithm Design**: A new algorithm using an efficient dual - module - calculation algorithm is proposed, and the complexity bound of this algorithm is given. 3. **Experimental Verification**: Experiments were carried out through Maple implementation, and the cases in which this algorithm is faster than existing methods were discussed. ### Mathematical Expression Let \(\ell = (\ell_\alpha)_{\alpha \in \mathbb{N}^n}\) be an \(n\)-dimensional sequence, where \(\ell_\alpha\in K\) represents an element in the field \(K\). Given a polynomial \(f(X)=\sum_{\gamma\in B\subseteq\mathbb{N}^n}f_\gamma X^\gamma\in\text{ann}(\ell)\), by definition we have: \[ \sum_{\gamma\in B\subseteq\mathbb{N}^n}f_\gamma\ell_{\gamma + \alpha}=0,\quad\forall\alpha\in\mathbb{N}^n \] Since the sequence \(\ell\) gives values only on its finite support set \(S_\ell\), for all \(\alpha\notin S_\ell\) we have \(\ell_\alpha = 0\), so the above equation can be transformed into a finite linear equation system with variables \(f_\gamma\). ### Hankel Matrix To represent this linear equation system, a Hankel matrix \(H_\ell\) can be constructed, whose size is \(s := |S_\ell|\). The columns of the matrix \(H_\ell\) are labeled by the monomials in \(S_\ell\) in ascending order, and the rows are labeled by \(X^\alpha\cdot\ell\), where \(\alpha\in S_\ell\). The \((\alpha,\beta)\)-th element of the matrix \(H_\ell\) is: \[ (H_\ell)_{\alpha,\beta}=(X^\beta\cdot\ell)_\alpha=\ell_{\alpha + \beta} \] Therefore, \(H_\ell\) is a Hankel matrix, and its complexity can be bounded by \(O(s^\omega)\), where \(\omega\) is the complexity constant of matrix multiplication. ### Key Idea of the New Algorithm The core idea of the new algorithm is to use the calculation method of the dual module to solve the recurrence relations. Specifically, the author introduces the concept of "reciprocity" of sequences and applies it to the transformed dual - module calculation. By this method, the size of the matrix required for calculation can be effectively reduced, thereby improving the efficiency of the algorithm. ### Summary The main objective of this paper is to provide a more efficient method for calculating the linear recurrence relations of multi - dimensional sequences. By transforming the problem into the calculation of the orthogonal space in the dual module and through experimental verification, the effectiveness of this method is verified.