GitHub - auto-differentiation/xad: Powerful automatic differentiation in C++ and Python (original) (raw)

🚀 XAD: Powerful Automatic Differentiation for C++ & Python

XAD is the ultimate solution for automatic differentiation, combining ease of use with high performance. It's designed to help you differentiate complex applications with speed and precision—whether you're optimizing neural networks, solving scientific problems, or performing financial risk analysis.

🌟 Why XAD?

XAD is trusted by professionals for its speed, flexibility, and scalability across various fields:

Machine Learning & Deep Learning: Accelerate neural network training and model optimization.
Optimization in Engineering & Finance: Solve complex problems with high precision.
Numerical Analysis: Improve methods for solving differential equations efficiently.
Scientific Computing: Simulate physical systems and processes with precision.
Risk Management & Quantitative Finance: Assess and hedge risks in sophisticated financial models.
Computer Graphics: Optimize rendering algorithms for high-quality graphics.
Robotics: Enhance control and simulation for robotic systems.
Meteorology: Improve accuracy in weather prediction models.
Biotechnology: Model complex biological processes effectively.

Key Features

Forward & Adjoint Mode: Supports any order using operator overloading.
Vector modes: Compute multiple derivatives at once.
Checkpointing Support: Efficient tape memory management for large-scale applications.
External Function Interface: Seamlessly connect with external libraries.
Thread-Safe Tape: Ensure safe, concurrent operations.
High Performance: Optimized for speed and efficiency.
Proven in Production: Battle-tested in large-scale, mission-critical systems.
Exception-Safe: Formal guarantees for stability and error handling.
Eigen support: Works with the popular linear algebra library Eigen

💻 Example

Calculate first-order derivatives of an arbitrary function with two inputs and one output using XAD in adjoint mode.

Adouble x0 = 1.3; // initialise inputs Adouble x1 = 5.2; tape.registerInput(x0); // register independent variables tape.registerInput(x1); // with the tape tape.newRecording(); // start recording derivatives Adouble y = func(x0, x1); // run main function tape.registerOutput(y); // register the output variable derivative(y) = 1.0; // seed output adjoint to 1.0 tape.computeAdjoints(); // roll back adjoints to inputs cout << "dy/dx0=" << derivative(x0) << "\n" << "dy/dx1=" << derivative(x1) << "\n";

🚀 Getting Started

git clone https://github.com/auto-differentiation/xad.git cd xad mkdir build cd build cmake .. make

For more detailed guides, refer to our Installation Guideand explore Tutorials.

🤝 Contributing

Want to get involved? We welcome contributors from all backgrounds! Check out our Contributing Guide and join the conversation in ourDiscussions.

🐛 Found a Bug?

Please report any issues through ourIssue Tracker.

XAD-Py: XAD in Python.
QuantLib-Risks: Fast risk evaluations in C++ and Python.