pylops

Matrix-free linear operators enable you to solve large-scale inverse problems without forming explicit matrices by defining forward and adjoint operations directly. This approach ensures scalable operator composition and inversion for complex geoscience workflows.

Seismic deconvolution using matrix-free operators is best handled by defining a forward operator, providing the seismic data, and running a solver like LSQR or NormalEquations. This workflow recovers the unknown signal efficiently without explicit matrix storage.

Yes, you can use FISTA and RegularizedInversion for tomography applications. These solvers support optional regularization to robustly invert matrix-free operators, enabling stable subsurface imaging and general operator-based modelling in geoscience.

Matrix-free operator algebra fully supports stacking and composing operators. You can construct forward and adjoint operators, stack them vertically or horizontally, and combine them to build complex inversion pipelines for large-scale inverse problems.

When solving inverse problems with matrix-free operators, limitations include the need to correctly implement both forward and adjoint operations for custom operators, and convergence dependencies on solver choice and regularization parameters.

Yes, you need numpy and matplotlib as dependencies to run matrix-free linear operator solvers. Numpy handles the numerical array computations for operator algebra, while matplotlib provides visualization for evaluating solver results and inverted signals.