ControlParameters

Last modified 4 weeks ago

Control Parameters for Minimizer

max_threads

• maximal number of threads to start

proportional_stop | absolute_stop | absolute_score | absolute_iter | absolute_time

• type of stopping criterion and the value of error

stop_count

• number of steps for stopping

population_size

• number of members of the population

recombination_strategy = de_3_exp_rand | de_3_bin_rand | de_3_exp_rand_T | de_3_bin_rand_T | simple | de_3_bin | de_3_exp | de_3_exp_T | de_3_bin_T

• type of evolution

recombination_weight

• scaling factor, 0 means adaptation

recombination_prob

• probability, 0 means adaptation

recombination_gamma

• adaptation parameter

es_lambda

• number of individuals to exchange between branches.

es_cutoff

• =3 number of individuals

es_kind

• with what to substitute: 0 - best, 1 - less better

mean_cost

• =0

noglobal_eps

• small or big cube in parameter space, 0, -1 - big and initial guess is ignored.

substeps

* 0 - use limits, -1 - do not perturb, 0.1 - use vecinity

substieps

* 0 - use limits, -1 - do not perturb, 0.1 - use vecinity

ca_flag

* 0 - use space for the best, 1 - distribute equaly

logdepth

* log more info

transform = tanh | sin | alg | rand

• type of transform for parameters with box constraints

gamma_init

• parameter for constraints transform

roundoff_error

• safety offset

seed

• seed for pseudo random numbers

step_parameter

• the scaling coefficient for gradient step.

step_decrement

• the coefficient to decrease the gradient step if the previous one wasn't successful.

derivative_step

• the parameter to compute the derivative numerically.

number_of_trials

• number of times to decrease the gradient step before giving up.

run_before

• functions to run befor main loop

run

• functions to run in main loop with number of tau's

run_after

• functions to run after main loop

Example:

[default_settings]
max_threads=2
proportional_stop=1e-04
stop_count=15
absolute_iter=10
population_size=150
recombination_strategy=de_3_exp_rand
recombination_weight=0
recombination_prob=0
recombination_gamma=0.9
es_lambda=25
es_cutoff=3
es_kind=0
mean_cost=0
noglobal_eps=0
substeps=0.005
substieps=0
ca_flag=1
logdepth=2
transform=tanh
gamma_init=1.0
precision=16
keep_order=1
roundoff_error=1e-06
seed=62534198
run_before=initcancel;optpost;optposteval;
run=gcadeep;5;gacdeep;1;gdeep;1;edeep;1;sdeep;24;selde;1;dpupdate;1;checkcancel;1;substitute;2;optpost;1;optposteval;1;
run_after=optpost;optposteval;printlog;

Control Parameters for the Quality Functional

debug=0 | 1

ignore_cost=0 | 1

constrain_aggr=sum | max

penalty_aggr=sum | max

<name>=<keyword>;<keyword>;<index>;<rank>;<weight>

<weight> is the multiplier for the function value in the final value

<rank> is the value with wich the random number is compared to decide if the decrease in this function is worth

mainfunc=target;objfunc;0;1;1; - scoring function. 0 means it is the main scoring function

func01=penalty;readpenalty;1;0.45;1; - read the value of the penalty from scoring program output

func02=penalty;readpenalty;2;0.45;1;

Example:

[default_target]
debug=0
ignore_cost=0
constrain_aggr=sum
penalty_aggr=sum
mainfunc=target;objfunc;0;1;0;
cbd_pen_em=penalty;readpenalty;1;2;1;
cbd_pen_r1=penalty;readpenalty;2;2;1;
cbd_pen_r3=penalty;readpenalty;3;2;0;
cbd_pen_r5=penalty;readpenalty;4;2;0;
cbd_pen_r7=penalty;readpenalty;5;2;0;
cbd_pen_r8=penalty;readpenalty;6;2;0;