Resumen

JABBOUR, George  y  PAREDES, Jose Luis. Analysis of the efficiency of weighted median-based signal reconstruction algorithm through Cox regression. uct [online]. 2012, vol.16, n.64, pp.173-180. ISSN 1316-4821.

In this paper, the efficiency of the algorithm for compressive sensing (CS) signal reconstruction based on weighted median regression (WMR) is analyzed through a Cox-regression model. We perform 1620 reconstructions for signals with different dimension (N), sparsity (K), number of measurements (M) and regularization parameter (α) that induces sparsity in the solution. Among the most relevant results, we find that the algorithm efficiency, as a function of the regularization parameter, follows an inverted parabolic function reaching its maximum at α = 0.8. Furthermore, we show that the reconstruction algorithm is quite sensible to α and M. Thus, a slight change on those parameters leads to a notable variation on the algorithm’s convergence speed. Therefore, by suitably tunning the number of measurements, we can control the volatile described above. Thus, if the ratio N/M goes from 7 to 9, the probability of having a good performance increases from 0.4 to 0.7. Furthermore, if α changes from 0.5 to 0.9 this probability increases from 0.14 to 0.96.

Palabras clave : Signal Reconstruction; Compressive Sensing; Weighted Median; Cox Regression; Survival Analysis.

        · resumen en Español     · texto en Español     · Español ( pdf )