Fig. 10

2-algorithm at low attenuation: log-likelihoods (LL) of (μ,λ) (with respect to the scattered and true data), and normalized mean square errors (NMSE) of μ and λ, respectively, during the first 100 sub-iterations (5 iterations) of the low-attenuation 2-algorithm. Note that the trues-MLEM activity updates [sub-iterations 0 to 10, 20 to 30, etc.] are supposed to increase the trues LL, explaining decreases in the scatter LLs, and vice versa