Code to reproduce the results of the paper

A.J. Weinstein and M.B. Wakin. Online Search Orthogonal Matching Pursuit. In IEEE Statistical Signal Processing Workshop.

• Python >=2.6
• NumPy >= 1.3
• SciPy >= 0.7
• Matplotlib >= 1.2

Optional (to reproduce the A*OMP results)

• Matlab
• mlabwrap v1.1-pre
• A*OMP

If you want to run A*OMP, make sure to add the matlab directory to Matlab (e.g., by adding addpath path/to/osomp/matlab line to startup.m.

Create residue comparison plot (Fig. 2 of the paper)

$ python osomp.py --residue

Create rate of recovery plot (Fig. 3(a-c))

$ python osomp.py --rate

To also run A*OMP:

$ python osomp.py --rate --astar

Create relative error for noisy observations (Fig. 3(d))

$ python osomp.py --noisy