Algebraic Framework for Discrete Dynamical Systems over Laurent Series
Ramamonjy Andriamifidisoa, Loukman Ben Saindou · 2025
Read Algebraic Framework for Discrete Dynamical Systems over Laurent Series by Ramamonjy Andriamifidisoa, Loukman Ben Saindou (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-04708.txt.
Introduction
Ramamonjy Andriamifidisoa and Loukman Ben Saindou extend an algebra framework in 2025 to systems that move in both directions over time. Their base is Laurent series, which allow negative as well as positive powers, a natural fit for two way sequences. An older framework for discrete algebraic systems worked only with one way forms.
They rebuild the two key spaces of that framework, add a pairing map, and extend the shift operator. The duality and adjoint traits of the older theory are proved to survive the move. Examples and a data processing case study on two way sequence changes show the machinery at work. The authors answer an open question from the earlier work and point to uses in image filtering and control theory for multi dimensional systems.
Catalog record
| CREATOR | Ramamonjy Andriamifidisoa, Loukman Ben Saindou |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-04708
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Algebraic Framework for Discrete Dynamical Systems over Laurent Series Ramamonjy Andriamifidisoa, Loukman Ben Saindou Abstract . We generalize the framework of discrete algebraic dynamical systems [ 1 ] to Laurent polynomials and series over ℤ r \Z^{r} , enabling the modeling of bidirectional discrete systems. By redefining the spaces 𝐃 ′ \mathbf{D}^{\prime} and 𝐀 ′ \mathbf{A}^{\prime} , introducing a bilinear mapping (defined as the scalar product in Section 3), and extending the shift operator, we preserve the duality and adjoint properties of [ 1 ] . These properties are rigorously proved and illustrated through examples and a data processing case study on bidirectional sequence transformations. In contrast to Oberst [ 3 ] , our algebraic approach emphasizes the structure of Laurent series, providing a streamlined framework for multidimensional systems. This work addresses an open question from [ 1 ] and has applications in multidimensional data processing, such as image filtering and control theory. Keywords : Discrete algebraic dynamical systems, Laurent series, shift operator, duality, multidimensional systems. 1. Introduction Discrete algebraic dynamical systems, as developed in [ 1 ] , model multidimensional systems using polynomials in \mathbf{D}=\F[X_{1},\dots,X_{r}] and sequences in \mathbf{A}=\F^{\N^{r}} , indexed by \N^{r} . Our work also builds on recent developments in discrete linear dynamical systems [ 2 ] and the algebraic theory of multidimensional systems [ 4 ] . Extending the time set from \N^{r} to \Z^{r} enables modeling bidirectional systems, such as two-sided filters in data processing and control systems with infinite memory. This work extends [ 1 ] to Laurent polynomials in \mathbf{D}^{\prime}=\F[X_{1},X_{1}^{-1},\dots,X_{r},X_{r}^{-1}] and Laurent series in \mathbf{A}^{\prime}=\F[[Y_{1},Y_{1}^{-1},\dots,Y_{r},Y_{r}^{-1}]] , corresponding to \Z^{r} . (0,0) \alpha_{1} \alpha_{2} (a) \mathbb{N}^{2} (0,0) \alpha_{1} \alpha_{2} (b) \mathbb{Z}^{2} Figure 1. Comparaison of \mathbb{N}^{r} and \mathbb{Z}^{r} Indexing Grids. (a) Unidirectional grid \mathbb{N}^{2} (for r=2 ), showing non-negative indices (\alpha_{1},\alpha_{2})\in\mathbb{N}^{2} . Points (in blue) are restricted to the positive quadrant, suitable for causal systems (e.g., unidirectional image filtering). (b) Bidirectional grid \mathbb{Z}^{2} , including negative indices, allowing bidirectional transformations (e.g., non-causal filters with infinite memory). Our extension preserves the duality and adjoint properties of [ 1 ] . Unlike Oberst [ 3 ] , who focuses on constant systems, we emphasize the algebraic structure of Laurent series, with explicit constructions for a bilinear mapping (defined as the scalar product in Section 3) and shift operator, supported by examples and a case study. We prove these properties hold in \Z^{r} . Applications include multidimensional data processing, such as image filtering, and control theory. To the best of our knowledge, no recent works have explored this algebraic approach for bidirectional systems over \Z^{r} . The paper is organized as follows: Section 2 introduces definitions with examples, Section 3 presents results on duality, the shift operator, and a case study, and Section 4 concludes with future directions. 2. Definitions and Notations 2.1. Notations Let \N denote the natural numbers, \Z the integers, \F a commutative field, and r\geq 1 . The set \Z^{r} is the r -dimensional integer lattice, serving as the time set for discrete systems. • Sequence Spaces : – \F^{\Z^{r}}=\{W:\Z^{r}\to\F\} , the vector space of sequences W=(W_{\alpha})_{\alpha\in\Z^{r}} , with W(\alpha)=W_{\alpha}\in\F . – \F^{(\Z^{r})}=\{W\in\F^{\Z^{r}}\mid W_{\alpha}\neq 0\text{ for finitely many }\alpha\} , the subspace of sequences with finite support. Both \F^{\Z^{r}} and \F^{(\Z^{r})} are \F -vector spaces under pointwise addition and scalar multiplication. For \alpha=(\alpha_{1},\dots,\alpha_{r})\in\Z^{r} , define monomials X^{\alpha}=X_{1}^{\alpha_{1}}\cdots X_{r}^{\alpha_{r}} , Y^{\alpha}=Y_{1}^{\alpha_{1}}\cdots Y_{r}^{\alpha_{r}} , with X_{i},X_{i}^{-1},Y_{i},Y_{i}^{-1} formal variables. For \alpha\in\Z^{r} , let \delta_{\alpha}:\Z^{r}\to\F be: \delta_{\alpha}(\beta)=\begin{cases}1&\text{if }\alpha=\beta,\\ 0&\text{otherwise}.\end{cases} 2.2. Spaces Definition 2.1 (Spaces) . Let \F be a commutative field and r\geq 1 . Define: (1) \mathbf{D}^{\prime}=\F[X_{1},X_{1}^{-1},\dots,X_{r},X_{r}^{-1}] , the ring of Laurent polynomials, with elements: d(X)=\sum_{\alpha\in\Z^{r}}d_{\alpha}X^{\alpha},\quad d_{\alpha}\in\F,\quad\text{finitely many }d_{\alpha}\neq 0. (2) \mathbf{A}^{\prime}=\F[[Y_{1},Y_{1}^{-1},\dots,Y_{r},Y_{r}^{-1}]] , the space of Laurent series, with elements: W(Y)=\sum_{\alpha\in\Z^{r}}W_{\alpha}Y^{\alpha},\quad W_{\alpha}\in\F. Remark 2.2 (Sequences and Signals) . A sequence is an element W(Y)=\sum_{\alpha\in\Z^{r}}W_{\alpha}Y^{\alpha}\in\mathbf{A}^{\prime} , where W_{\alpha}\in\F are coefficients indexed by \Z^{r} . In applications, such sequences are often called signals , representing discrete data indexed by time or space (e.g., audio samples for r=1 or image pixels for r=2 ). We use “sequence” to emphasize the algebraic structure, reserving “signal” for applied contexts. Example 2.3 (Laurent Polynomials) . (1) For r=1 , the Laurent polynomial d(X)=5X^{-1}-3X^{2} has coefficients d_{-1}=5 , d_{2}=-3 , and d_{k}=0 for all other k\in\Z . (2) For r=2 , an example is d(X_{1},X_{2})=X_{1}^{-1}X_{2}+3X_{1}^{2}X_{2}^{-2} , with non-zero coefficients at \alpha=(-1,1) and \alpha=(2,-2) . Example 2.4 (Laurent Series) . For r=1 , a Laurent series is W(Y)=\dots+2Y^{-1}+Y^{1}+3Y^{2}+\dots . Example 2.5 (Unidirectional vs. Bidirectional Indexing) . For r=1 , consider a sequence W(Y)=\sum_{n\in\mathbb{Z}}W_{n}Y^{n} . • In \mathbb{N} (unidirectional), W_{n}=0\;\text{for}\;n<0 , representing a causal sequence (e.g., audio samples starting at t=0 ). • In \mathbb{Z} (bidirectional), W_{n} can be non-zero for n<0 , enabling bidirectional transformations, such as a smoothing filter P(X)=0.5X^{-1}+0.5X , where W^{\prime}_{n}=0.5(W_{n-1}+W_{n+1}) , using past and future values (see Example 3.8 ). Proposition 2.6 ( Isomorphisms) . Let \F be a commutative field and r\geq 1 . The following isomorphisms hold : (1) \mathbf{D}^{\prime}\simeq\F^{(\Z^{r})} , via the map X^{\alpha}\leftrightarrow\delta_{\alpha} . (2) \mathbf{A}^{\prime}\simeq\F^{\Z^{r}} , via the map W(Y)\leftrightarrow(W_{\alpha})_{\alpha\in\Z^{r}} . Proof. For (1), the map \phi:\mathbf{D}^{\prime}\to\mathbb{F}^{\left(\mathbb{Z}^{r}\right)} , defined by \phi(X^{\alpha})=\delta_{\alpha} , extends linearly to d(X)=\sum_{\alpha\in\mathbb{Z}^{r}}d_{\alpha}X^{\alpha}\mapsto\sum_{\alpha\in\mathbb{Z}^{r}}d_{\alpha}\delta_{\alpha} . Since d(X) has finitely many non-zero d_{\alpha} , the image lies in \mathbb{F}^{\left(\mathbb{Z}^{r}\right)} . The map is bijective, as each \delta_{\alpha} corresponds uniquely to X^{\alpha} . For (2), the map \psi:\mathbf{A}^{\prime}\to\mathbb{F}^{\left(\mathbb{Z}^{r}\right)} , defined by \psi(W(Y))=(W_{\alpha})_{\alpha\in\mathbb{Z}^{r}} , where W(Y)=\sum_{\alpha\in\mathbb{Z}^{r}}W_{\alpha}Y^{\alpha} , is bijective, as it associates each Laurent series with its sequence of coefficients, with inverse mapping (W_{\alpha})_{\alpha\in\mathbb{Z}^{r}}\mapsto\sum_{\alpha\in\mathbb{Z}^{r}}W_{\alpha}Y^{\alpha} . ∎ 2.3. Discrete Linear Dynamical Systems Definition 2.7 ( General Discrete Linear Dynamical System) . Following Willems [ 8 , 9 , 10 ] , a discrete linear dynamical system over \mathbb{Z}^{r} is a triplet \Sigma=(\mathbb{T},\mathbb{W},\mathcal{B}) , where : • \mathbb{T}=\mathbb{Z}^{r} is the time set, • \mathbb{W}=\mathbb{F}^{l} is the signal space ( an \mathbb{F} -vector space), • \mathcal{B}\subseteq(\mathbf{A}^{\prime})^{l} is the behavior, a \mathbf{D}^{\prime} -submodule representing the set of admissible signal trajectories. Remark 2.8 . The behavioral approach of Willems [ 8 , 9 , 10 ] provides a general framework for modeling dynamical systems, where the behavior \mathcal{B} describes all possible trajectories of the system without distinguishing inputs and outputs. This flexibility allows modeling a wide range of systems, including those not necessarily defined by linear equations. Example 2.9 ( Periodic Trajectories) . For r=1,\;\mathbb{F}=\mathbb{R},\;\mathbb{T}=\mathbb{Z},\;\mathbb{W}=\mathbb{R} , and \mathcal{B}=\{W\in\mathbb{A}^{\prime}\mid W_{n+2}=W_{n},\forall n\in\mathbb{Z}\} . This behavior describes sequences that are periodic with period 2. For instance, a valid trajectory is W(Y)=\sum_{n\in\mathbb{Z}}a_{n}Y^{n},\;\text{where}\;a_{n}=1\;\text{if}\;n is even and a_{n}=0 if n is odd, i.e., W(Y)=\sum_{n\;\text{even}}Y^{n} . This system is linear and shift-invariant but not necessarily defined by a Laurent polynomial kernel. Example 2.10 (Bounded Trajectories) . For r=2,\;\mathbb{F}=\mathbb{R},\;\mathbb{T}=\mathbb{Z}^{2},\;\mathbb{W}=\mathbb{R} and \mathcal{B}=\{W\in\mathbb{A}^{\prime}\mid|W_{\alpha}|\leq 1,\forall\;\alpha\in\mathbb{Z}^{2}\} . This behavior consists of all sequences ( or signals, e.g, image pixel intensities) whose values are bounded by 1. An example trajectory is W(Y_{1},Y_{2})=\sum_{\alpha\in\mathbb{Z}^{2}}W_{\alpha}Y^{\alpha} , where W_{\alpha}=\sin(\alpha_{1}+\alpha_{2}) . This system illustrates the generality of willems’ framework, as \mathcal{B} is not defined by a polynomial equation but is still a valid \mathbf{D}^{\prime} -submodule. Definition 2.11 (Autoregressif Systems) . A discrete linear dynamical system is called autoregressive if its behavior is defined as : \mathcal{B}=\ker R(X)=\{W(Y)=(W_{1}(Y),\ldots,W_{l}(Y))\in(\mathbf{A}^{\prime})^{l}\mid R(X)\circ W(Y)=0\}, where R(X)\in(\mathbf{D}^{\prime})^{k,l} is a matrix of Laurent polynomials, and the shift operator \circ is given by : R(X)\circ W(Y)=\left(\sum_{j=1}^{l}R_{ij}(X)\circ W_{j}(Y)\right)_{1\leq