| Contents |
| We discuss the application of weighted and re-weighted l1 minimization for recovering sparse signals whose support set is known to have been drawn according to a non-uniform prior distribution over s-sparse support sets. In particular, we show that in this regime, weighted l1 minimization can outperform unweighted l1 minimization in terms of number of measurements needed to achieve a given reconstruction accuracy. Finally, we leverage this theory to provide recovery guarantees for reweighted l1 minimization as an effective tool for dynamic
filtering to track time-varying sparse signals. |
|