Theoretica

‍A C++ math library for scientific computing with a simple and elegant interface.

Written in modern C++, Theoretica is a comprehensive suite of numerical methods designed for high-performance scientific computing, focusing on elegance and simplicity, lowering the learning curve typically associated with such software.

Across many applications in science and data analysis, Theoretica offers the speed and robustness of C++ with the readability of Python.

If you'd like to join us, to learn or to bring your expertise, make sure to read the Onboarding Guide and the Contributing Guide. Your help is valuable!

Why Theoretica?

• 🧠 Intuitive API: Mathematical expressions should look like math, not like machine code.
• ⚡ Blazing Fast: Built with zero-overhead abstractions to squeeze out every drop of performance from your CPU.
• 🔋 Batteries Included: Out-of-the-box support for advanced numerical methods, without any dependencies.
• 🖥️ HPC Support: Growing support for hardware accelleration, making your simulations even faster.
• 🎚️ Embedded Friendly: Runs anywhere, even with low resources, making it a great choice for embedded systems.

Theoretica is constantly developed and improved with new ideas!

Features

Theoretica handles the heavy lifting across multiple scientific domains:

• Solvers: Differential equation solvers, Monte Carlo methods, integral quadratures.
• Linear Algebra: Standalone linear algebra algorithms for dense matrices.
• Automatic Differentiation: Evaluate exact derivatives without manual analytical calculations or numerical approximations.
• Optimization: Built-in routines for finding function extrema and roots, in arbitrary dimensions.
• Statistics: Methods for regression, fits, histograms, data sampling, and built-in error propagation mechanics.
• Signal Processing: Standalone Fast Fourier Transform support.
• File I/O: First-class support for storing large-scale data sets using CSV and HDF5 file formats, as well as Data Frames.

There's also a new experimental GUI module being developed, for visualizing results directly in your code.

Showcase: Simulating Chaos

The following code solves a differential equation, such as the Lorenz attractor:

vec3 f(real t, vec3 v) {
 
    const real a = 13, b = 20, c = 8/3.0;
    const real x = v[0], y = v[1], z = v[2];
 
    return {
        a * (y - x),
        x * (b - z) - y,
        x * y - c * z
    };
}
 
int main() {
 
    // Solve the system using Runge-Kutta's method
    vec3 v0 = {0.1, 0, 0};
    auto solution = ode::solve_rk4(f, v, 0.0, 50.0);
 
    // Write the solution directly to file
    std::ofstream file ("attractor.csv");
    file << solution;
}

You can find many more examples in the examples folder.

Setup

Theoretica is a header-only library and has no dependencies, so you can include it in your projects straight-away! To use the library, you can include single headers or use theoretica.h which includes all modules, or alternatively include theoretica_mini.h which includes only base modules.

You can also compile tests and example programs using Make (make test and make examples) or using CMake:

cd build
cmake ..
# Use your chosen compiler

Documentation

The documentation for the project is available here. The documentation is written using Doxygen syntax alongside the source code and the online version is automatically updated on each commit. The bibliography used during research for the library is listed in the Bibliography. To learn more about the design choices behind the library, you can read the RFC documents in this folder.

Join Us!

We believe that the best scientific software is built by communities. Regardless of your background, your participation is valuable.

How you can make an impact:

• 🐛 Report Bugs: Found a bug? Open an issue!
• 💡 Suggest Features: Need a specific numerical method or feature? Just let us know.
• ⌨️ Write Code: Check out our good first issue labels. We are always looking for new solvers, optimizations and features!
• 📖 Improve Docs: Help us make Theoretica accessible to scientists and students worldwide.

Have a look at the Contributing Guide to learn more!

License

The project is currently under the GNU Lesser General Public License 3.0. You may learn more about it here.

---