run() parameters

Main parameters

Parameter Name Type Default Documentation
energy_window (float,float) Estimated lower and upper bounds of the spectrum. Negative values of the lower bound will be reset to 0 for susceptibilities and conductivity.
max_time int -1 = infinite Maximum runtime in seconds, use -1 to set infinite.
verbosity int 2 on MPI rank 0, 0 otherwise. Verbosity level (max level - 3).
t int 50 Number of elementary updates per global update (\(T\)). Bigger values make the algorithm more ergodic.
f int 100 Number of global updates (\(F\)); ignored if adjust_f = True. Bigger values make the algorithm more ergodic.
adjust_f bool False Adjust the number of global updates automatically.
l int 2000 Number of particular solutions used in the final accumulation (\(L\)); ignored if adjust_l = True. Bigger values reduce noise in the final solution / make it smoother.
adjust_l bool False Adjust the number of solutions used in the final accumulation.
make_histograms bool False Accumulate histograms of objective function values.

Fine tuning options

Parameter Name Type Default Documentation
random_seed int 34788 + 928374 * MPI.rank Seed for random number generator.
random_name str “” Name of random number generator.
max_rects int 60 Maximum number of rectangles to represent spectra (\(K_{max}\)), should be below 70.
min_rect_width float 1e-3 Minimal width of a rectangle, in units of the energy window width.
min_rect_weight float 1e-3 Minimal weight of a rectangle, in units of the requested solution norm.
distrib_d_max float 2 Maximal parameter of the power-law distribution function for the Metropolis algorithm.
gamma float 2 Proposal probability parameter \(\gamma\).
adjust_f_range (int,int) (100,5000) Search range for the number of global updates.
adjust_f_l int 20 Number of particular solutions used to adjust \(F\).
adjust_f_kappa float 0.25 Limiting value of \(\kappa\) used to adjust \(F\).
adjust_l_range (int,int) (100,2000) Search range for the number of solutions used in the final accumulation.
adjust_l_good_d float 2.0 Maximal ratio \(D/D_\mathrm{min}\) for a particular solution to be considered good.
adjust_l_verygood_d float 4/3 Maximal ratio \(D/D_\mathrm{min}\) for a particular solution to be considered very good.
adjust_l_ratio float 0.95 Critical ratio \(N_\mathrm{very good}/N_\mathrm{good}\) to stop \(L\)-adjustment procedure.
hist_max float 2.0 Right boundary of the histograms, in units of \(D_\mathrm{min}\) (left boundary is always set to \(D_\mathrm{min}\)).
hist_n_bins int 100 Number of bins for the histograms.