APEMoST
Defines

src/define_defaults.h File Reference

Go to the source code of this file.

Defines

#define N_BETA   20
#define BETA_0   -0.001
#define ITER_LIMIT   100000
#define MUL   0.85
#define N_SWAP   -30
#define PARAMS_FILENAME   "params"
#define DATA_FILENAME   "data"
#define TARGET_ACCEPTANCE_RATE   0.50
#define MAX_AR_DEVIATION   0.01

Define Documentation

#define BETA_0   -0.001

Position where the smallest/hottest chain should start. Remember, beta = 1/T, so beta = 0 is infinitely hot.

Has to be between 0 and 1. If < 0, it is determined automatically so that the hottest chain will aim at a stepwidth of 1/3 of parameter space.

Referenced by calibrate_rest(), and check().

#define DATA_FILENAME   "data"

Referenced by check(), help_phase(), main(), and setup_chains().

#define ITER_LIMIT   100000

How many iterations should be used for the burn-in phase? How many iterations should be used after burn-in for adjusting the step widths?

Referenced by calibrate_first(), calibrate_rest(), and check().

#define MAX_AR_DEVIATION   0.01

How much deviation from the desired acceptance rate is acceptable

Referenced by calibrate_first(), calibrate_rest(), and check().

#define MUL   0.85

Factor used for scaling the step widths

Referenced by calibrate_first(), calibrate_rest(), and check().

#define N_BETA   20

Number of chains to use for parallel tempering

<= 12 is not recommended

Referenced by analyse_data_probability(), analyse_marginal_distributions(), calibrate_rest(), check(), prepare_and_run_sampler(), and setup_chains().

#define N_SWAP   -30

After how many iterations should a swap occur?

Performance evaluations suggest to set this to 2000/N_BETA. If < 0, this will be done for you.

Referenced by check(), and prepare_and_run_sampler().

#define PARAMS_FILENAME   "params"
#define TARGET_ACCEPTANCE_RATE   0.50

Desired acceptance rate after calibration

Referenced by adapt(), calibrate_first(), calibrate_rest(), check(), markov_chain_calibrate_orig(), and rmw_adapt_stepwidth().