|
HYPE
|
|
HYPE
|
Public Data Type Components | |
| logical | task_mc = .FALSE. |
| logical | task_boundps = .FALSE. |
| logical | task_scanning = .FALSE. |
| logical | task_brentnew = .FALSE. |
| logical | task_stpstdesc = .FALSE. |
| logical | task_dfp = .FALSE. |
| logical | task_bfgs = .FALSE. |
| logical | task_demc = .FALSE. |
| logical | task_stagemc = .FALSE. |
| logical | task_runens = .FALSE. |
| logical | task_ensall = .FALSE. |
| logical | task_ensbest = .FALSE. |
| logical | task_writeall = .FALSE. |
| logical | task_writesim = .FALSE. |
| integer | scan_xpoints = 1 |
| integer | scan_ypoints = 1 |
| integer | nruns_mc = 1000 |
| integer | nruns_mci = 200 |
| integer | nruns_mcimax = 100 |
| integer | nruns_best = 1 |
| integer | nruns_stages = 1 |
| real | nruns_zoom = 0.9 |
| logical | cal_log = .TRUE. |
| integer | cal_debugcase = 0 |
| integer | cal_maxiterat = 100 |
| real | cal_maxtime = 72 |
| integer | cal_improvparamiter = 10 |
| integer | cal_improvcrititer = 10 |
| real | cal_improvcrittol = 0.001 |
| real | qn_flattol = 0.001 |
| integer | qn_epsiltype = 1 |
| real | qn_factorderiv = 0.02 |
| integer | qn_stencil = 2 |
| real | qn_lambdamaxfac = 0.90 |
| real | qn_lambdaaccel = 1.618 |
| logical | brent_diagonalstep = .TRUE. |
| integer | linesearch_maxiter = 500 |
| real | linesearch_tol = 0.001 |
| integer | demc_ngen = 100 |
| integer | demc_npop = 25 |
| real | demc_gammascale = 1. |
| real | demc_crossover = 1. |
| real | demc_sigma = 0.1 |
| real | demc_accprob = 0. |
| integer | nruns_simloop = 1 |
Type for holding information about optimization and parameter ensemble simulation.
| logical worldvar::optimizationtype::brent_diagonalstep = .TRUE. |
Flag to take a diagonal step at the end of each Brent iteration.
| integer worldvar::optimizationtype::cal_debugcase = 0 |
Flag to indicate debug case.
| integer worldvar::optimizationtype::cal_improvcrititer = 10 |
Amount of last optimisation iterations taken into account for criteria improvement monitoring.
| real worldvar::optimizationtype::cal_improvcrittol = 0.001 |
Tolerance to consider criteria as optimised (delta/mean over the last iterations)
| integer worldvar::optimizationtype::cal_improvparamiter = 10 |
Amount of last optimisation iterations taken into account for parameter improvement monitoring.
| logical worldvar::optimizationtype::cal_log = .TRUE. |
Flag to write all calibration details to file "calibration.log" (Y/N flag; Y = .TRUE., N = .FALSE.)
| integer worldvar::optimizationtype::cal_maxiterat = 100 |
Max amount of allowed iterations.
| real worldvar::optimizationtype::cal_maxtime = 72 |
Max amout of time (hours) allowed to calibration routine.
| real worldvar::optimizationtype::demc_accprob = 0. |
Probability of acceptance of a worse propose that is very close to currently best.
| real worldvar::optimizationtype::demc_crossover = 1. |
Crossover probability (default=1)
| real worldvar::optimizationtype::demc_gammascale = 1. |
Multiplicative scaling of the mutation strength gamma = 2.38/(2*num_par^2)^0.5.
| integer worldvar::optimizationtype::demc_ngen = 100 |
Number of generations.
| integer worldvar::optimizationtype::demc_npop = 25 |
Number of populations.
| real worldvar::optimizationtype::demc_sigma = 0.1 |
Sample Error Standard deviation.
| integer worldvar::optimizationtype::linesearch_maxiter = 500 |
Maximum amount of iterations allowed for line search algorithm.
| real worldvar::optimizationtype::linesearch_tol = 0.001 |
Tolerance for line search of minimum [Fred].
| integer worldvar::optimizationtype::nruns_best = 1 |
Number of ensembles = number of best runs saved from random simulations.
| integer worldvar::optimizationtype::nruns_mc = 1000 |
Number of runs for Monte Carlo simulation.
| integer worldvar::optimizationtype::nruns_mci = 200 |
Number of runs for each iteration reducing the parameter space.
| integer worldvar::optimizationtype::nruns_mcimax = 100 |
Maximum number of iteration for reducing the parameter space.
| integer worldvar::optimizationtype::nruns_simloop = 1 |
Number of parameter ensembles = number of best runs saved from random simulations.
| integer worldvar::optimizationtype::nruns_stages = 1 |
Number of stages to repeat num_mc runs, centering around the best runs at each stage [Fred, 26.08.10].
| real worldvar::optimizationtype::nruns_zoom = 0.9 |
Zooming factor for the centering [Fred, 26.08.10].
| integer worldvar::optimizationtype::qn_epsiltype = 1 |
Specifies whether to use absolute (case 1, default), relative (case 2) or mixed epsilon (case 0) values.
| real worldvar::optimizationtype::qn_factorderiv = 0.02 |
Factor to offset current parameter value for numerical derivative.
| real worldvar::optimizationtype::qn_flattol = 0.001 |
Tolerance for gradient norm to be considered zero (quasi-Newton)
| real worldvar::optimizationtype::qn_lambdaaccel = 1.618 |
Factor increasing the step length proposed by QN algorithms (case lambda = 1 to be replaced by lambdaAccel), in order to allow for faster iteration progression; default value consistent with golden ratio line search algorithm.
| real worldvar::optimizationtype::qn_lambdamaxfac = 0.90 |
Factor to contain lambda prior to line search (allows to leave space between current point and boundary, for numerical derivatives)
| integer worldvar::optimizationtype::qn_stencil = 2 |
Numerical derivative stencil type.
| integer worldvar::optimizationtype::scan_xpoints = 1 |
Number of points taken for 1st parameter.
| integer worldvar::optimizationtype::scan_ypoints = 1 |
Number of points taken for 2nd parameter.
| logical worldvar::optimizationtype::task_bfgs = .FALSE. |
Quasi-Newton gradient-based calibration with BFGS algorithm for inverse Hessian update.
| logical worldvar::optimizationtype::task_boundps = .FALSE. |
Repeated MonteCarlo simulation with reduced parameter space.
| logical worldvar::optimizationtype::task_brentnew = .FALSE. |
Calibration using the Brent method (2010 version with new line search)
| logical worldvar::optimizationtype::task_demc = .FALSE. |
DEMC Differential-Evolution Markov Chain (Monte Carlo) simulation.
| logical worldvar::optimizationtype::task_dfp = .FALSE. |
Quasi-Newton gradient-based calibration with DFP algorithm for inverse Hessian update.
| logical worldvar::optimizationtype::task_ensall = .FALSE. |
Running ensembles based on allsims parameters.
| logical worldvar::optimizationtype::task_ensbest = .FALSE. |
Running ensembles basid on bestsims parameters.
| logical worldvar::optimizationtype::task_mc = .FALSE. |
MonteCarlo simulation.
| logical worldvar::optimizationtype::task_runens = .FALSE. |
Running ensembles and write results from ensemble runs.
| logical worldvar::optimizationtype::task_scanning = .FALSE. |
Scan mode, for 2-param situations only.
| logical worldvar::optimizationtype::task_stagemc = .FALSE. |
MonteCarlo simulations using stage-wise centering around set of bests MC runs.
| logical worldvar::optimizationtype::task_stpstdesc = .FALSE. |
Steepest descent, implemented as a particular case of quasi-Newton gradient-based calibration.
| logical worldvar::optimizationtype::task_writeall = .FALSE. |
Write performance for all simulations from ensemble runs.
| logical worldvar::optimizationtype::task_writesim = .FALSE. |
Write simulations results for all simulations from ensemble runs (MC-methods only)
1.9.4
