
i
i
i
i
i
i
i
i
13.5 Experiments on Real Data Sets 289
Figure 13.12 Results on temperature data from the Pacific Ocean.
the total coefficients. Although the rest of the coefficients are not strictly zero, we
may set K = 40.
The compressive sensing theory suggests that data can be reconstructed with high
probability from M = 3K −4K random measurements. Figure 13.12c shows the re-
construction performance with different numbers of random measurements. Each
indicated data point is averaged over 10 test runs to avoid fluctuations. Apparently,
the reconstruction precision increases as M increases. A steep rise is observed in
both figures when M becomes greater than