Skip to main content
. 2013 Jun;3(3):147–161. doi: 10.3978/j.issn.2223-4292.2013.06.01

Figure 2.

Figure 2

Equi-contour lines corresponding to the L2, L1, L0 norm cost functions in the case where Inline graphic is a 2D vector. The L1, L0 norms have a property that they yield larger costs to non-sparse points located far from the coordinate axes, which leads to a strong power in picking up sparse solutions better than the L2 norm