Some stochastic process techniques applied to deterministic models
Eric José Ávila-Vales, José Villa-Morales · 2025
Read Some stochastic process techniques applied to deterministic models by Eric José Ávila-Vales, José Villa-Morales (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-2508-02700.txt.
Introduction
Eric José Ávila-Vales and José Villa-Morales bring chance tools to fixed-rule models in a 2025 paper on chance maths. Chance models help read and call hard acts in the world, the pair notes in the text. Their work reads when a chance run quits a bound zone in its path through space.
Mean quit time and the spread of quit times are the twin reads sought in the work. Two kinds of spread laws yield those reads when solved in the frame at hand. A free code scheme runs the maths in two and three spans for real use.
Catalog record
| CREATOR | Eric José Ávila-Vales, José Villa-Morales |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 54 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-02700
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Some stochastic process techniques applied to deterministic models Eric José Ávila-Vales, José Villa-Morales Abstract Stochastic mathematical models are essential tools for understanding and predicting complex phenomena. The purpose of this work is to study the exit times of a stochastic dynamical system—specifically, the mean exit time and the distribution of exit times of the stochastic process within a bounded domain. These quantities are obtained by solving elliptic and parabolic partial differential equations (PDEs), respectively. To support practical applications, we propose a numerical scheme implemented in FreeFEM, emphasizing its effectiveness in two- and three-dimensional cases due to the software’s limitations in higher dimensions. The examples provided illustrate the theoretical results, which extend known one-dimensional solutions to higher-dimensional settings. This contribution bridges theoretical and computational approaches for analyzing stochastic processes in multidimensional domains, offering insights into their behavior and potential applications. Keywords: Partial differential equations, stochastic models, exit mean time. MSC [2020] 35B09, 60G53, 92B05, 65M06 1 Introduction Currently, many decision‐making processes rely on mathematical models, underscoring the importance of their rigorous study. A well‐designed mathematical model aims to describe the phenomenon of interest as comprehensively as possible, incorporating only those parameters and variables that have a significant impact and omitting extraneous factors that would render the model impractical (see [ 1 ] ). On the other hand, as is often the case in time‐dependent models, the system’s conditions may change from one day to the next. Consequently, the predictions of a purely deterministic model can become unreliable or irrelevant (see [ 2 , 3 ] ). A common—and sometimes effective—remedy is to adopt a model that accounts for such variations, including the uncertainty arising from imprecise parameter measurements (see [ 4 ] ). Thus, if we assume that a deterministic model captures the “average” behavior of a phenomenon, then, in general terms, a stochastic model can be viewed as accounting for unpredictable perturbations or effects that are difficult to include explicitly. Our first step, therefore, is to construct a stochastic counterpart. There are multiple methods to achieve this (for instance, see [ 5 ] ). When the underlying dynamics are governed by a system of ordinary differential equations, a standard approach is to posit that, over small time intervals, the system undergoes random state transitions with prescribed probabilities (see [ 6 , 7 ] ). In Section 2 , we demonstrate—in a generic example—how to derive a stochastic dynamical system from a deterministic one (see Theorem 1 ). Once we have the stochastic model, we use stochastic calculus techniques to study some functionals of interest. Namely, given a region of particular interest (D), where the process starts, we determine the average time the process takes to leave that region (Theorem 3 ) and, similarly, we find the distribution of that exit time (Theorem 5 ). These quantities depend on the solution of certain Partial Differential Equations (PDEs), the first of an Elliptic Equation ( 8 ) and the second of a Parabolic Equation ( 12 ), see [ 8 , 9 ] . The next step is to propose a solution scheme for these equations, see Section 4 . Since our purpose is for the theoretical results to be applied by a broad academic sector, we undertook the task of investigating the various software available to solve PDEs. Among them, some are paid and others are free access, the vast majority. We looked for, among other software characteristics, that it had certain previous versions to ensure the stability and quality of the results, that it was easy to code, and that it was used by a large community to ensure the existence of support forums. Among them, we chose FreeFEM, version 4.13, for being free software and meeting the aforementioned characteristics. Moreover, it is worth noting that some artificial intelligences know the FreeFEM language, which helps in coding. It is important to mention that, of all the software consulted, none of them solve PDE problems in dimensions greater than 4. There are two reasons for this: the first is that most applications focus on physics, and the second is that as dimensions increase, numerical methods quickly lose precision. Therefore, if a specific problem of interest arises in higher dimensions, a numerical method from the many that exist can be applied, see [ 10 ] . The theoretical results considered in this work can be framed in any dimension; however, the limitation of the software to solve PDEs in dimensions no greater than 4 means that our exposition is limited to dimensions 2 and 3. The case of dimension 1 is well known, see [ 11 ] , which is why we have omitted it. In summary, our contribution is as follows: given a stochastic dynamical system and a specified starting region, we employ partial differential equations and numerical‐analysis techniques to study the mean exit time and its distribution from that region. To the best of our knowledge, these results are well established in one dimension but have not been derived—or at least not implemented—for higher‐dimensional settings. To demonstrate the effectiveness of our approach, Section 5 presents illustrative examples from the literature that showcase the insights provided by our theoretical findings. The work is organized as follows. In Section 2 , we introduce the stochastic model and show that it has a solution, Y . If Y(0)\in D , where D is a bounded domain, in Section 3 , we determine the expected mean exit time of Y from the set D and its distribution. In Section 4 , the numerical scheme is discussed, and the FreeFEM code is generally described, see [ 12 ] . Finally, in Section 5 , several examples taken from the literature, are presented, and their solutions are discussed. 2 Some Stochastic Models As noted in Section 1, the results we develop can be applied broadly to the solution of any SDE. As a representative example, we employ the construction method described in [ 7 ] to derive a stochastic model. Let us consider a phenomenon whose dynamic Y(t)=(y_{1}(t),y_{2}(t),y_{3}(t))^{T} is governed by the following system of differential equations \displaystyle\frac{dy_{1}}{dt} \displaystyle= \displaystyle f_{1}(y_{1})-c_{1,1}g_{1}(y_{1},y_{2})-c_{1,2}g_{2}(y_{1},y_{3}), \displaystyle\frac{dy_{2}}{dt} \displaystyle= \displaystyle f_{2}(y_{2})-c_{2,1}g_{1}(y_{1},y_{2})-c_{2,2}g_{3}(y_{2},y_{3}), \displaystyle\frac{dy_{3}}{dt} \displaystyle= \displaystyle f_{3}(y_{3})-c_{3,1}g_{3}(y_{2},y_{3})-c_{3,2}g_{2}(y_{1},y_{3}), where c_{i,j}\geq 0 are real constants, and f_{i}:\mathbb{R}\to\mathbb{R} , g_{i}:\mathbb{R}\times\mathbb{R}\to\mathbb{R} are non-negative continuous functions. The components of the deterministic system can be described as follows: • The function f_{i}(y_{i}) represents the growth rate of the variable y_{i} , independent of other variables. • The function g_{n}(y_{i},y_{j}) represents the reduction in the growth rate of the variable y_{m} due to the influence of the variables y_{i} and y_{j} . This contribution is weighted by the constant c_{i,j} . Suppose that over a small time interval \Delta t , the vector \Delta Y can take only the following seven states: \displaystyle(1,0,0)^{T},\quad(0,1,0)^{T},\quad(0,0,1)^{T},\quad(-c_{1,1},-c_{2,1},0)^{T}, \displaystyle(-c_{1,2},0,-c_{3,2})^{T},\quad(0,-c_{2,2},-c_{3,1})^{T},\quad(0,0,0)^{T}. The probabilities associated with these states are shown in Table 1 . Change Probability \Delta Y^{1}=(1,0,0)^{T} p_{1}:=f_{1}(y_{1})\Delta t \Delta Y^{2}=(0,1,0)^{T} p_{2}:=f_{2}(y_{2})\Delta t \Delta Y^{3}=(0,0,1)^{T} p_{3}:=f_{3}(y_{3})\Delta t \Delta Y^{4}=(-c_{1,1},-c_{2,1},0)^{T} p_{4}:=g_{1}(y_{1},y_{2})\Delta t \Delta Y^{5}=(-c_{1,2},0,-c_{3,2})^{T} p_{5}:=g_{2}(y_{1},y_{3})\Delta t \Delta Y^{6}=(0,-c_{2,2},-c_{3,1})^{T} p_{6}:=g_{3}(y_{2},y_{3})\Delta t \Delta Y^{7}=(0,0,0)^{T} p_{7}:=1-\sum_{i=1}^{6}p_{i} Table 1 : Possible changes with their corresponding probabilities. From this, the drift vector is given by B=\left(\begin{array}[]{c}b_{1}\\ b_{2}\\ b_{3}\end{array}\right):=\frac{E[\Delta Y]}{\Delta t}=\frac{1}{\Delta t}\sum_{i=1}^{7}p_{i}\Delta Y^{i}=\left(\begin{array}[]{c}f_{1}-c_{1,1}g_{1}-c_{1,2}g_{2}\\ f_{2}-c_{2,1}g_{1}-c_{2,2}g_{3}\\ f_{3}-c_{3,1}g_{3}-c_{3,2}g_{2}\end{array}\right), and the diffusion matrix is \displaystyle A=(a_{i,j}) \displaystyle:= \displaystyle\frac{E[\Delta Y(\Delta