DifferentialEquations.jl: Scientific Machine Learning (SciML) Enabled Simulation and Estimation
This is a suite for numerically solving differential equations written in Julia and available for use in Julia, Python, and R. The purpose of this package is to supply efficient Julia implementations of solvers for various differential equations. Equations within the realm of this package include:
- Discrete equations (function maps, discrete stochastic (Gillespie/Markov) simulations)
- Ordinary differential equations (ODEs)
- Split and Partitioned ODEs (Symplectic integrators, IMEX Methods)
- Stochastic ordinary differential equations (SODEs or SDEs)
- Stochastic differential-algebraic equations (SDAEs)
- Random differential equations (RODEs or RDEs)
- Differential algebraic equations (DAEs)
- Delay differential equations (DDEs)
- Neutral, retarded, and algebraic delay differential equations (NDDEs, RDDEs, and DDAEs)
- Stochastic delay differential equations (SDDEs)
- Experimental support for stochastic neutral, retarded, and algebraic delay differential equations (SNDDEs, SRDDEs, and SDDAEs)
- Mixed discrete and continuous equations (Hybrid Equations, Jump Diffusions)
- (Stochastic) partial differential equations ((S)PDEs) (with both finite difference and finite element methods)
The well-optimized DifferentialEquations solvers benchmark as some of the fastest implementations, using classic algorithms and ones from recent research which routinely outperform the "standard" C/Fortran methods, and include algorithms optimized for high-precision and HPC applications. At the same time, it wraps the classic C/Fortran methods, making it easy to switch over to them whenever necessary. Solving differential equations with different methods from different languages and packages can be done by changing one line of code, allowing for easy benchmarking to ensure you are using the fastest method possible.
DifferentialEquations.jl integrates with the Julia package sphere with:
- GPU acceleration through CUDA.jl and DiffEqGPU.jl
- Automated sparsity detection with SparsityDetection.jl
- Automatic Jacobian coloring with SparseDiffTools.jl, allowing for fast solutions to problems with sparse or structured (Tridiagonal, Banded, BlockBanded, etc.) Jacobians
- Allowing the specification of linear solvers for maximal efficiency
- Progress meter integration with the Juno IDE for estimated time to solution
- Automatic plotting of time series and phase plots
- Built-in interpolations
- Wraps for common C/Fortran methods like Sundials and Hairer's radau
- Arbitrary precision with BigFloats and Arbfloats
- Arbitrary array types, allowing the definition of differential equations on matrices and distributed arrays
- Unit checked arithmetic with Unitful
Additionally, DifferentialEquations.jl comes with built-in analysis features, including:
- Forward and Adjoint Sensitivity Analysis (Automatic Differentiation) for fast gradient computations
- Parameter Estimation and Bayesian Analysis
- Neural differential equations with DiffEqFlux.jl for efficient scientific machine learning (scientific ML) and scientific AI.
- Automatic distributed, multithreaded, and GPU Parallel Ensemble Simulations
- Global Sensitivity Analysis
- Uncertainty Quantification
If you have any questions, or just want to chat about solvers/using the package, please feel free to use the Gitter channel. For bug reports, feature requests, etc., please submit an issue. If you're interested in contributing, please see the Developer Documentation.
Supporting and Citing
The software in this ecosystem was developed as part of academic research. If you would like to help support it, please star the repository as such metrics may help us secure funding in the future. If you use SciML software as part of your research, teaching, or other activities, we would be grateful if you could cite our work.
@article{rackauckas2017differentialequations,
title={Differentialequations.jl--a performant and feature-rich ecosystem for solving differential equations in julia},
author={Rackauckas, Christopher and Nie, Qing},
journal={Journal of Open Research Software},
volume={5},
number={1},
year={2017},
publisher={Ubiquity Press}
}
is necessary for any use of DifferentialEquations.jl or the packages that are maintained as part of its suite (OrdinaryDiffEq.jl, Sundials.jl, DiffEqDevTools.jl, etc.). Additionally, many of the solvers utilize novel algorithms, and if these algorithms are used we asked that you cite the methods. Please see our citation page for guidelines.
Getting Started: Installation And First Steps
Installing from Julia
To install the package, use the following command inside the Julia REPL:
using Pkg
Pkg.add("DifferentialEquations")
To load the package, use the command:
using DifferentialEquations
This will add solvers and dependencies for all kinds of Differential Equations (e.g. ODEs or SDEs etc., see the Supported Equations section below). If you are interested in only one type of equation solvers of DifferentialEquations.jl
or simply want a more lightweight version, see the Low Dependency Usage page.
To understand the package in more detail, check out the following tutorials in this manual. It is highly recommended that new users start with the ODE tutorial. Example IJulia notebooks can also be found in DiffEqTutorials.jl. If you find any example where there seems to be an error, please open an issue.
For the most up to date information on using the package, please join the Gitter channel.
Using the bleeding edge for the latest features and development is only recommended for power users. Information on how to get to the bleeding edge is found in the developer documentation.
Installing from Python
Use of DifferentialEquations.jl from the Python programming language is available through the diffeqpy module. To install diffeqpy, use pip:
pip install diffeqpy
Using diffeqpy requires that Julia is installed and in the path, along with DifferentialEquations.jl and PyCall.jl. To install Julia, download a generic binary from the JuliaLang site and add it to your path. To install Julia packages required for diffeqpy, open up Python interpreter then run:
>>> import diffeqpy
>>> diffeqpy.install()
and you're good! In addition, to improve the performance of your code it is recommended that you use Numba to JIT compile your derivative functions. To install Numba, use:
pip install numba
diffeqpy supports the majority of DifferentialEquations.jl with very similar syntax, see the diffeqpy README for more details. One important point to note is that Numba is generally an order of magnitude slower than Julia in terms of the generated differential equation solver code, and thus it is recommended to use julia.Main.eval
for Julia-side derivative function implementations for maximal efficiency. See this blog post for more information.
Installing from R
Use of DifferentialEquations.jl from the R programming language is available through the diffeqr module. diffeqr is registered into CRAN. Thus to add the package, use:
install.packages("diffeqr")
To install the master branch of the package (for developers), use:
devtools::install_github('SciML/diffeqr', build_vignettes=T)
You will need a working installation of Julia in your path. To install Julia, download a generic binary from the JuliaLang site and add it to your path. The download and installation of DifferentialEquations.jl will happen on the first invocation of diffeqr::diffeq_setup()
.
Currently, use from R supported a subset of DifferentialEquations.jl which is documented through CRAN.
IJulia Notebook Tutorials
You can access extra tutorials supplied in the DiffEqTutorials.jl repository via the commands:
using Pkg
pkg"add https://github.com/SciML/SciMLTutorials.jl"
using SciMLTutorials
SciMLTutorials.open_notebooks()
Or you can view the webpages for the rendered tutorials at the links found in the repository.
Video Tutorial
Tutorials
The following tutorials will introduce you to the functionality of DifferentialEquations.jl. More examples can be found by checking out the IJulia notebooks in the examples folder.
- Ordinary Differential Equations
- Stochastic Differential Equations
- Delay Differential Equations
- Differential Algebraic Equations
- Continuous-Time Jump Processes and Gillespie Methods
- Illustrative Model: SIR disease dynamics
- Defining the SIR Model using Reactions via Catalyst
- Building and Simulating the Jump Process from Catalyst Models
- Building and Simulating the Jump Process using the DiffEqJump Low-level Interface
- Defining the Jumps Directly:
ConstantRateJump
- SSAStepper
- Reducing Memory Use: Controlling Saving Behavior
- Defining the Jumps Directly:
MassActionJump
- Defining the Jumps Directly: Mixing
ConstantRateJump
andMassActionJump
- Adding Jumps to a Differential Equation
- Adding a VariableRateJump
- RegularJumps and Tau-Leaping
- FAQ
- Jump Diffusion Equations
- Boundary Value Problems
- Additional Tutorials
Removing and Reducing Compile Times
In some situations one may wish to decrease the compile time associated with DifferentialEquations.jl usage. If that's the case, there's two strategies to employ. One strategy is to use the low dependency usage. DifferentialEquations.jl is a metapackage composed of many smaller packages, and thus one could directly use a single component, such as OrdinaryDiffEq.jl
for the pure Julia ODE solvers, and decrease the compile times by ignoring the rest (note: the interface is exactly the same, except using a solver other than those in OrdinaryDiffEq.jl will error). We recommend that downstream packages only rely on exactly the packages they need.
The other strategy is to use PackageCompiler.jl to create a system image that precompiles the whole package. To do this, one simply does:
using PackageCompiler
PackageCompiler.create_sysimage([:DifferentialEquations,:Plots];replace_default=true)
Note that there are some drawbacks to adding a package in your system image, for example the package will never update until you manually rebuild the system image again. For more information on the consequences, see this portion of the PackageCompiler manual
Basics
These pages introduce you to the core of DifferentialEquations.jl and the common interface. It explains the general workflow, options which are generally available, and the general tools for analysis.
- Overview of DifferentialEquations.jl
- Common Solver Options (Solve Keyword Arguments)
- Solution Handling
- Plot Functions
- Integrator Interface
- Problem Interface
- Frequently Asked Questions
- Solver Compatibility Chart
Problem Types
These pages describe building the problem types to define differential equations for the solvers, and the special features of the different solution types.
- Discrete Problems
- ODE Problems
- Dynamical, Hamiltonian and 2nd Order ODE Problems
- Split ODE Problems
- Steady State Problems
- BVP Problems
- SDE Problems
- RODE Problems
- DDE Problems
- DAE Problems
- Jump Problems
Solver Algorithms
These pages describe the solvers and available algorithms in detail.
- Discrete Solvers
- ODE Solvers
- Dynamical, Hamiltonian, and 2nd Order ODE Solvers
- Split ODE Solvers
- Steady State Solvers
- BVP Solvers
- Jump Problem and Jump Diffusion Solvers
- SDE Solvers
- RODE Solvers
- DDE Solvers
- DAE Solvers
- Solver Benchmarks
Additional Features
These sections discuss extra performance enhancements, event handling, and other in-depth features.
- Jacobians, Gradients, etc.
- DiffEq-Specific Array Types
- DiffEqOperators
- Noise Processes
- Specifying (Non)Linear Solvers and Preconditioners
- Event Handling and Callback Functions
- Callback Library
- Parallel Ensemble Simulations
- I/O: Saving and Loading Solution Data
- Low Dependency Usage
- Progress Bar Integration
Analysis Tools
Because DifferentialEquations.jl has a common interface on the solutions, it is easy to add functionality to the entire DiffEq ecosystem by developing it to the solution interface. These pages describe the add-on analysis tools which are available.
- ParameterizedFunctions
- Parameter Estimation and Bayesian Analysis
- Bifurcation Analysis
- Local Sensitivity Analysis (Automatic Differentiation)
- Global Sensitivity Analysis
- Uncertainty Quantification
- Neural Networks
- Algorithm Development and Testing
Modeling Tools
While DifferentialEquations.jl can be used to directly build any differential or difference equation (/ discrete stochastic) model, in many cases it can be helpful to have a tailored-built API for making certain types of common models easier. This is provided by the modeling functionality.
Extra Details
These are just assorted extra explanations for the curious.
Acknowledgements
Core Contributors
JuliaDiffEq and DifferentialEquations.jl has been a collaborative effort by many individuals. Significant contributions have been made by the following individuals:
- Chris Rackauckas (@ChrisRackauckas) (lead developer)
- Yingbo Ma (@YingboMa)
- David Widmann (@devmotion)
- Hendrik Ranocha (@ranocha)
- Ethan Levien (@elevien)
- Tom Short (@tshort)
- @dextorious
- Samuel Isaacson (@isaacsas)
Google Summer of Code Alumni
- Yingbo Ma (@YingboMa)
- Shivin Srivastava (@shivin9)
- Ayush Pandey (@Ayush-iitkgp)
- Xingjian Guo (@MSeeker1340)
- Shubham Maddhashiya (@sipah00)
- Vaibhav Kumar Dixit (@Vaibhavdixit02)