Low complexity approximate cyclic adaptive matching pursuit

Based on the iterated cyclic adaptive matching pursuit algorithm, we construct a low complexity approximate variant for finding sparse solutions to systems of linear equations. We employ a greedy neighbor permutation strategy coupled with an approximate scalar product matrix to ensure that the complexity of the algorithm remains low. The sparse solution is cyclically updated improving the performance while the sparsity level is estimated online using the predictive least squares criterion. The performance of the algorithm is similar to that of the non approximate variants while the complexity can be considerably lower.