Skip to main content
. Author manuscript; available in PMC: 2011 Feb 25.
Published in final edited form as: Stat Comput. 2011 Jan 4;21(2):261–273. doi: 10.1007/s11222-009-9166-3

Table 7.

Comparison of various algorithms for the PET imaging problem. A 4,096-pixel image is recovered from photon coincidence counts collected from 2,016 detector tubes. Here the roughness constant is μ = 10−6, and the convergence criterion is ε = 10−8. The number of parameters is 4,096

Algorithm Evals Objective Time
EM 3376 −15432.61 6836.05
q = 1 740 −15432.61 1743.55
q = 2 608 −15432.60 1432.38
q = 3 406 −15432.60 955.30
q = 4 372 −15432.57 875.51
q = 5 268 −15432.58 627.00
q = 6 222 −15432.57 520.97
q = 7 204 −15432.56 477.02
q = 8 188 −15432.54 441.48
q = 9 178 −15432.52 417.63
q = 10 176 −15432.51 411.20
q = 15 184 −15432.62 430.94
q = 20 236 −15432.45 559.32
SqS1 314 −15435.72 742.90
SqS2 290 −15432.54 684.79
SqS3 232 −15432.53 549.06