Identifying Differential Equations

[Lorin Searing]

Themost common classification of differential equations is based on order. The order of a differential equation simply is the order of its highest derivative. You can have first-, second-, and higher-order differential equations.

In order to identify a nonhomogeneous differential equation, you first need to know what a homogeneous differential equation looks like. You also often need to solve one before you can solve the other.

Nonhomogeneous differential equations are the same as homogeneous differential equations, except they can have terms involving only x (and constants) on the right side, as in this equation:

If you need to find particular solutions to nonhomogeneous differential equations, then you can start with the method of undetermined coefficients. Suppose you face the following nonhomogeneous differential equation:

Where can I find the most up-to-date or whatever you consider to be the most useful symmetry-finding package for differential equations? I do not intend to restrict to, but would like to include those, that are designed to work within Mathematica.

In Mathematics, a differential equation is an equation that contains one or more functions with its derivatives. The derivatives of the function define the rate of change of a function at a point. It is mainly used in fields such as physics, engineering, biology and so on. The primary purpose of the differential equation is the study of solutions that satisfy the equations and the properties of the solutions. Learn how to solve differential equations here.

One of the easiest ways to solve the differential equation is by using explicit formulas. In this article, let us discuss the definition, types, methods to solve the differential equation, order and degree of the differential equation, ordinary differential equations with real-word examples and a solved problem.

A differential equation is an equation which contains one or more terms and the derivatives of one variable (i.e., dependent variable) with respect to the other variable (i.e., independent variable)

A differential equation contains derivatives which are either partial derivatives or ordinary derivatives. The derivative represents a rate of change, and the differential equation describes a relationship between the quantity that is continuously varying with respect to the change in another quantity. There are a lot of differential equations formulas to find the solution of the derivatives.

You can see in the first example, it is a first-order differential equation which has degree equal to 1. All the linear equations in the form of derivatives are in the first order. It has only the first derivative such as dy/dx, where x and y are the two variables and is represented as:

A function that satisfies the given differential equation is called its solution. The solution that contains as many arbitrary constants as the order of the differential equation is called a general solution. The solution free from arbitrary constants is called a particular solution. There exist two methods to find the solution of the differential equation.

Differential equations have several applications in different fields such as applied mathematics, science, and engineering. Apart from the technical applications, they are also used in solving many real life problems. Let us see some differential equation applications in real-time.

The various other applications in engineering are: heat conduction analysis, in physics it can be used to understand the motion of waves. The ordinary differential equation can be utilized as an application in the engineering field for finding the relationship between various parts of the bridge.

To understand Differential equations, let us consider this simple example. Have you ever thought about why a hot cup of coffee cools down when kept under normal conditions? According to Newton, cooling of a hot body is proportional to the temperature difference between its temperature T and the temperature T0 of its surrounding. This statement in terms of mathematics can be written as:

1. An ordinary differential equation contains one independent variable and its derivatives. It is frequently called ODE. The general definition of the ordinary differential equation is of the form: Given an F, a function os x and y and derivative of y, we have

The different types of differential equations are:

Ordinary Differential Equations

Partial Differential Equations

Homogeneous Differential Equations

Non-homogeneous Differential Equations

Linear Differential Equations

Nonlinear Differential Equations

The order of the highest order derivative present in the differential equation is called the order of the equation. If the order of the differential equation is 1, then it is called the first order. If the order of the equation is 2, then it is called a second-order, and so on.

The main purpose of the differential equation is to compute the function over its entire domain. It is used to describe the exponential growth or decay over time. It has the ability to predict the world around us. It is widely used in various fields such as Physics, Chemistry, Biology, Economics and so on.

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Discovering nonlinear differential equations that describe system dynamics from empirical data is a fundamental challenge in contemporary science. While current methods can identify such equations, they often require extensive manual hyperparameter tuning, limiting their applicability. Here, we propose a methodology to identify dynamical laws by integrating denoising techniques to smooth the signal, sparse regression to identify the relevant parameters, and bootstrap confidence intervals to quantify the uncertainty of the estimates. We evaluate our method on well-known ordinary differential equations with an ensemble of random initial conditions, time series of increasing length, and varying signal-to-noise ratios. Our algorithm consistently identifies three-dimensional systems, given moderately-sized time series and high levels of signal quality relative to background noise. By accurately discovering dynamical systems automatically, our methodology has the potential to impact the understanding of complex systems, especially in fields where data are abundant, but developing mathematical models demands considerable effort.

Since Newton discovered the second law of motion, scientists have sought to formulate mathematical models in the form of differential equations that accurately represent natural phenomena. In the past half-century, dynamical systems have been employed in various disciplines such as physics1,2, chemistry3, biology4,5,6,7,8,9, neuroscience10,11,12, epidemiology13,14, ecology15,16 and environmental sciences17,18. Nonetheless, developing these models remains challenging and typically requires considerable effort from specialists in the relevant fields19,20.

As early as the 1980s, scientists turned to statistical methods to reverse engineer governing equations for nonlinear systems from data21. This approach, often referred to as the inverse problem22 or system identification23, aims to automatically discover mathematical models that accurately represent the inherent dynamics. Building on this foundation, symbolic regression has been instrumental in advancing our ability to develop more interpretable models of complex systems24,25. Sparse regression has emerged as a practical method for this problem, eliminating the time-consuming task of determining equations manually. A remarkable breakthrough is the sparse identification of nonlinear dynamics (SINDy)26, an approach that employs a sparsity-promoting framework to identify interpretable models from data by only selecting the most dominant candidate terms from a high-dimensional nonlinear-function space. This methodology has significantly advanced system identification, serving as a foundational influence for numerous subsequent sparse regression techniques27,28,29. Over time, SINDy has evolved and expanded its framework, incorporating Bayesian sparse regression30 and ensemble methods to estimate inclusion probabilities31, integrating neural networks32,33, and deploying tools to better manage noisy data34.

Among these advancements, a distinct variant, SINDy with AIC, aims to automate the model selection procedure35. This approach uses a grid of sparsity-promoting threshold parameters in conjunction with the Akaike information criterion (AIC) to determine the model that most accurately characterizes the dynamics of a given system. However, it encounters several obstacles that limit its practicality. Key challenges include its dependence on prior knowledge of the governing equations for model validation and identification, as well as the requirement for high-quality measurements given its limited capacity to compute numerical derivatives from previously unseen data. Furthermore, the efficacy of SINDy with AIC has only been demonstrated on data sets generated using specific initial conditions, sufficient observations, and low levels of noise, indicating the need for more comprehensive and rigorous analyses to assess its performance in diverse settings.

While existing methods, such as SINDy, use the Savitzky-Golay filter to both reduce noise and compute numerical derivatives, they require users to manually select the filter parameters36,37,38. Additionally, effective system identification often hinges on rigorous variable selection methods. To address these concerns, our contribution lies in developing an automated approach that employs a grid to fine-tune the Savitzky-Golay filter parameters and subsequently leverages bootstrapping to estimate confidence intervals and establish the governing terms of the system. As a result, our algorithm significantly improves the accuracy and efficiency of model discovery in low to medium-noise conditions while requiring only the assumptions of model sparsity and the presence of governing terms in the design matrix. We demonstrate the effectiveness of our approach by examining its success rate on synthetic data sets generated from known ordinary differential equations, exploring a range of initial conditions, time series of increasing length, and various noise intensities. Our algorithm automates the discovery of three-dimensional systems more efficiently than SINDy with AIC, achieving higher identification accuracy with moderately sized data sets and high signal quality.

3a8082e126