-
Notifications
You must be signed in to change notification settings - Fork 2
/
template.ini
77 lines (57 loc) · 4.29 KB
/
template.ini
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
####################################################################################################
# This is a template *ini file
# note:
# T = t = true = True
# F = f = false = False
#
# TODO: add error control option
####################################################################################################
chain_root = some_name # root of the names of the chains to be generated.
chain_num = 1 # how many chains will be generated. Note that in this version, all the chains
# are generated one after the other, so if the dimension of the problem is high,
# it is recommended to generate more chains, so that one can check then
# convergence by comparing the latest generated chains to those first generated
# chains.
walker_num = 200 # number of walkers in the ensemble.
burnin_step = 50 # how many times of iterations for burn-in.
sample_step = 1000 # this is actually the number of iterations, so the true size of the sample is
# " sample_step * walker_num "
skip_step = 10 # after one chain is finished (the sampling is still going), you can choose to
# skip some steps of writing samples into chain files.
# default value is 10.
efficient_a = 2.0 # this factor controls the acceptance ratio, when the dimension of the problem
# is low, 2 is usually a good choice. But when the dimension is high, it is better
# to set a smaller value, and MUST properly increase the sample_step, since the
# convergence rate is always becoming slow in high dimensional cases.
init_ball_radius = 0.5 # (reduced) radius of ND-ball, in which the sampler will randomly picks
# values to initialize the walkers. Note: this radius MUST > 0 and < 1
start_from_fiducial = true # if true, then the initial walkers are drawn around the fiducial values.
use_cosmomc_format = true # if true, then the sampled chains will be in the same format as that
# for CosmoMC::GetDist, however, in this case, the weights in the first
# columns are all set to 1.
# if false, then the weights will be set to \exp(-0.5*\chisq)
save_burned_ashes = true # if true, the burn-bin chains will be saved, from which you can see the
# 'evolution' of the chisq.
save_state_for_N_steps = 50
stop_on_error = false # whether stop when error happens (inside likelihood functions) ?
# set the ranges of your parameters
# fiducial_value:: this is actually used only for generating the first sampling position in
# parameter space. Lower_bound and upper_bound define the legal sampling range, if jump outside that
# range, then prior function will return 0, other it will always return 1. The fiducial_value should
# sit between lower_bound and upper_bound
param1 = fiducial_value1, lower_bound1, upper_bound1
param2 = fiducial_value2, lower_bound2, upper_bound2
####################################################################################################
# NOTE: names are case sensitive !!!
# the output order is exactly the same as you listed. If output_params is not set, then the output
# order might be as good as you wish, the order is determined by C++ map<> internally ... so I
# suggest you write down all the parameters of interests in the order you like.
####################################################################################################
output_params = param1 param2 ... # could be separated by space or comma
output_dparams = dparam1 dparam2 ... # same as output_params, but these are derived parameters
write_chain_header = false # whether write the parameter names as the heads of the chains. If you
# want to use GetDist.py to process the chains, it is better to set this
# to FALSE, if not errors might happen when loading chains.
# Update info: the error does not exist if pandas (0.17.0) is installed.
# ================================================================================================
# other settings you might wish to add, i.e., parameters to control data set usage