PyLops

Matrix-free linear operators for large-scale inverse problems.

An open-source, backend-agnostic, collection of composable linear operators and state-of-the-art solvers for practical scientific workflows.

GitHub Org.

Key features

Matrix-free operators

Represent operators by forward and adjoint matrix-vector products instead of explicit matrices.

Large-scale ready

Designed for problems where explicit matrices are prohibitive in memory and compute footprint.

State-of-the-art solvers

Includes iterative methods for least-squares and proximal optimization.

Extensible base operators

Provides LinearOperator and ProxOperator base classes inspired by SciPy and easy to extend.

Backend-agnostic design

Uses an idiomatic approach to provide support for multiple computational backends (NumPy / CuPy / JAX).

Open-source driven

Community-driven and affiliated with NumFOCUS, with strong focus on transparent development.

Domains

Maturity snapshot across key application areas.

Signal processing ★★★★★ Example
Image processing ★★★★★ Example
Medical imaging ★★★★★ Example
Geophysics ★★★★★ Example

Want to grow PyLops' competency in one of these domains, or add support for your own domain? Start a discussion here.

What we are up to

Open pull requests across our main repositories, ranked by latest activity.

Loading open pull requests…