Figure 2
Scatter plots for three examples of sources with estimations of the conditional expectation E [S 1 |S 2 ] and the linear regression given by ρS 2 : (a) corresponds to a pair of independent sources (sub-Gaussian); (b) corresponds to a pair of dependent sources with negative correlation following the model of abundances in spectral unmixing problem (constrained case); and (c) shows a pair of dependent sources generated using a Copula-t distribution with correlation ρ = 0.8. In (a) and (b) the LCE condition holds while in (c) it is only approximately verified.