1: \begin{abstract}

2: The convergence and numerical analysis of a low memory 

3: implementation of the Orthogonal Matching Pursuit 

4: greedy strategy, 

5:  which is termed Self Projected Matching Pursuit,

6: is presented. This approach provides an 

7:  iterative way of solving the least 

8: squares problem with much less storage requirement 

9: than direct linear algebra techniques. Hence, it is  

10:  appropriate for solving large linear systems.  Furthermore,   the low memory requirement of the method  

11: suits it for massive 

12: parallelization, via Graphics Processing Unit, to 

13: tackle systems which can be broken into a large 

14: number of subsystems of much smaller dimension. 

15: \end{abstract}

16: