Orthogonal Matching Pursuit (Python: NumPy + SciPy)

This is a simple implementation that uses scipy.optimize.nnls and numpy. It computes non-negative solutions by default, since that's what I needed it for, but can also be used to find unconstrained solutions by setting nonneg=False. The module favors convenience over performance, but performs reasonably well for many problems.

The Result object returned by omp is a self-contained expression of the problem that was solved and stores both the (optionally standardized) inputs, runtime parameters, details of the iteration, residual, and reconstructed signal.

The Jupyter notebook examples.ipynb contains step by step examples.