file diver_1.0.2/diver_1.0.2/diver.cpp
[No description available] More…
Namespaces
| Name |
|---|
| Gambit TODO: see if we can use this one: |
| Gambit::Diver_1_0_2 |
Functions
| Name | |
|---|---|
| scanner_plugin(diver , version(1, 0, 2) ) |
Detailed Description
Author: Pat Scott (p.scott@imperial.ac.uk)
Date: 2017 June
ScannerBit interface to Diver 1.0.2
Authors (add name and date if you modify):
Functions Documentation
function scanner_plugin
scanner_plugin(
diver ,
version(1, 0, 2)
)
=================================================
Interface to ScannerBit
Source code
// GAMBIT: Global and Modular BSM Inference Tool
// *********************************************
/// \file
///
/// ScannerBit interface to Diver 1.0.2
///
/// *********************************************
///
/// Authors (add name and date if you modify):
///
/// \author Pat Scott
/// (p.scott@imperial.ac.uk)
/// \date 2017 June
///
/// *********************************************
#include <vector>
#include <limits>
#include <fstream>
#include "gambit/ScannerBit/scanners/diver/1.0.2/diver.hpp"
#include "gambit/Utils/yaml_options.hpp"
#include "gambit/Utils/util_types.hpp"
#include "gambit/Utils/util_functions.hpp"
#include "gambit/Utils/variadic_functions.hpp"
/// =================================================
/// Interface to ScannerBit
/// =================================================
scanner_plugin(diver, version(1, 0, 2))
{
// Access Diver stuff and standard Gambit things
using namespace Gambit;
using namespace Gambit::Diver_1_0_2;
// Error thrown if the following entries are not present in the inifile
reqd_inifile_entries("NP");
// Tell cmake to search for the diver library.
reqd_libraries("diver");
// Set up the scan data container
diverScanData data;
// Code to execute when the plugin is loaded.
plugin_constructor
{
// Retrieve the external likelihood calculator
data.likelihood_function = get_purpose(get_inifile_value<std::string>("like"));
if (data.likelihood_function->getRank() == 0) cout << "Loading Diver differential evolution plugin for ScannerBit." << std::endl;
// Retrieve the external printer
data.printer = &(get_printer());
// Do not allow GAMBIT's own likelihood calculator to directly shut down the scan.
// Diver will assume responsibility for this process, triggered externally by
// the 'plugin_info.early_shutdown_in_progress()' function.
data.likelihood_function->disable_external_shutdown();
}
int plugin_main (void)
{
// Path to save Diver samples, resume files, etc
str defpath = get_inifile_value<str>("default_output_path");
str root = Utils::ensure_path_exists(get_inifile_value<str>("path",defpath+"Diver/native"));
// Ask the printer if this is a resumed run or not, and check that the necessary files exist if so.
bool resume = get_printer().resume_mode();
if (resume)
{
bool good = true;
static const std::vector<str> names = initVector<str>(root+".rparam", root+".devo", root+".raw");
for (auto it = names.begin(); it != names.end(); ++it)
{
std::ifstream file(*it);
good = good and file.good() and (file.peek() != std::ifstream::traits_type::eof());
file.close();
}
if (not good)
{
std::ostringstream warning;
warning << "Cannot resume previous Diver run because one or all of" << endl;
for (auto it = names.begin(); it != names.end(); ++it) warning << " " << *it << endl;
warning << "is missing or empty. This is probably because your last run didn't " << endl
<< "complete even one generation. Diver will start from scratch, " << endl
<< "as if you had specified -r.";
if (data.likelihood_function->getRank() == 0) cout << "WARNING: " << warning.str() << endl;
scan_warn << warning.str() << scan_end;
resume = false;
}
}
// Retrieve the global option specifying the minimum interesting likelihood.
double gl0 = get_inifile_value<double>("likelihood: model_invalid_for_lnlike_below");
// Retrieve the global option specifying the likelihood offset to use
double offset = get_inifile_value<double>("likelihood: lnlike_offset", 1e-4*gl0);
// Make sure the likleihood functor knows to apply the offset internally in ScannerBit
data.likelihood_function->setPurposeOffset(offset);
// Offset the minimum interesting likelihood by the offset, and flip it to match diver sign convention.
gl0 = -1.0 * (gl0 + offset);
// Other Diver run parameters
int nPar = get_dimension(); // Dimensionality of the parameter space
int nDerived = 0; // Number of derived quantities to output (GAMBIT printers handle these).
int nDiscrete = get_inifile_value<int> ("nDiscrete", 0); // Number of parameters that are to be treated as discrete
bool partitionDiscrete = get_inifile_value<bool> ("partitionDiscrete", false); // Split the population evenly amongst discrete parameters and evolve separately
int maxciv = get_inifile_value<int> ("maxciv", 1); // Maximum number of civilisations
int maxgen = get_inifile_value<int> ("maxgen", 5000); // Maximum number of generations per civilisation
int NP = get_inifile_value<int> ("NP"); // Population size (individuals per generation)
double Cr = get_inifile_value<double>("Cr", 0.9); // Crossover factor
double lambda = get_inifile_value<double>("lambda", 0.0); // Mixing factor between best and rand/current
bool current = get_inifile_value<bool> ("current", false); // Use current vector for mutation
bool expon = get_inifile_value<bool> ("expon", false); // Use exponential crossover
int bndry = get_inifile_value<int> ("bndry", 3); // Boundary constraint: 1=brick wall, 2=random re-initialization, 3=reflection
bool jDE = get_inifile_value<bool> ("jDE", true); // Use self-adaptive choices for rand/1/bin parameters as per Brest et al 2006
bool lambdajDE = get_inifile_value<bool> ("lambdajDE", true); // Use self-adaptive rand-to-best/1/bin parameters; based on Brest et al 2006
double convthresh = get_inifile_value<double>("convthresh", 1.e-3); // Threshold for gen-level convergence: smoothed fractional improvement in the mean population value
int convsteps = get_inifile_value<int> ("convsteps", 10); // Number of steps to smooth over when checking convergence
bool removeDuplicates = get_inifile_value<bool> ("removeDuplicates", true); // Weed out duplicate vectors within a single generation
bool doBayesian = false; // Calculate approximate log evidence and posterior weightings
double maxNodePop = 1.9; // Population at which node is partitioned in binary space partitioning for posterior
double Ztolerance = 0.01; // Input tolerance in log-evidence
int savecount = get_inifile_value<int> ("savecount", 1); // Save progress every savecount generations
bool native_output = get_inifile_value<bool> ("full_native_output", true); // Output .raw file (Diver native sample output format)
int init_pop_strategy = get_inifile_value<int> ("init_population_strategy", 2);// Initialisation strategy: 0=one shot, 1=n-shot, 2=n-shot with error if no valid vectors found.
int max_ini_attempts = get_inifile_value<int> ("max_initialisation_attempts", 10000); // Maximum number of times to try to find a valid vector for each slot in the initial population.
double max_acceptable_value= get_inifile_value<double>("max_acceptable_value",0.9999*gl0); // Maximum function value to accept for the initial generation if init_population_strategy > 0.
int verbose = get_inifile_value<int> ("verbosity", 0); // Output verbosity: 0=only error messages, 1=basic info, 2=civ-level info, 3+=population info
double (*prior)(const double[], const int, void*&) = NULL; // Pointer to prior function, only used if doBayesian = true.
void* context = &data; // Pointer to GAMBIT likelihood function and printers, passed through to objective function.
// Copy the contents of root to a char array.
std::vector<char> path(root.length()+1);
strcpy(&path[0], root.c_str());
// Unit cube boundaries
std::vector<double> lowerbounds(nPar, 0.0); // Lower boundaries of parameter space to scan
std::vector<double> upperbounds(nPar, 1.0); // Upper boundaries of parameter space to scan
// Scale factors
std::vector<double> Fvec = get_inifile_value<std::vector<double> >("F", initVector<double>(0.7));
int nF = Fvec.size(); // Size of the array indicating scale factors
// Discrete parameters
std::vector<int> discrete(nDiscrete, 0); // Indices of discrete parameters, Fortran style, i.e. starting at 1!!
//TODO Needs to be set automatically somehow? Not yet sure how to deal with discrete parameters in GAMBIT.
// Run Diver
if (data.likelihood_function->getRank() == 0) cout << "Starting Diver run..." << std::endl;
cdiver(&objective, nPar, &lowerbounds[0], &upperbounds[0], &path[0], nDerived, nDiscrete,
&discrete[0], partitionDiscrete, maxciv, maxgen, NP, nF, &Fvec[0], Cr, lambda, current,
expon, bndry, jDE, lambdajDE, convthresh, convsteps, removeDuplicates, doBayesian,
prior, maxNodePop, Ztolerance, savecount, resume, native_output, init_pop_strategy,
max_ini_attempts, max_acceptable_value, context, verbose);
if (data.likelihood_function->getRank() == 0) cout << "Diver run finished!" << std::endl;
return 0;
}
}
/// =================================================
/// Function definitions
/// =================================================
namespace Gambit
{
namespace Diver_1_0_2
{
//Function to be minimized. Corresponds to -ln(Likelihood). Redirects to the target of context pointer.
double objective(double params[], const int param_dim, int &fcall, bool &quit, const bool validvector, void*& context)
{
// Return the worst possible likelihood if the point is outside the prior box.
if (not validvector) return std::numeric_limits<double>::max();
// Put the parameters into a C++ vector
std::vector<double> param_vec(params, params + param_dim);
// Retrieve the likelihood function from the context pointer and call it
diverScanData* data = static_cast<diverScanData*>(context);
double lnlike = data->likelihood_function(param_vec);
// Increment the number of function calls, tell Diver to continue and return the likelihood
fcall += 1;
// Check whether the calling code wants us to shut down early
quit = Gambit::Scanner::Plugins::plugin_info.early_shutdown_in_progress();
return -lnlike;
}
}
}
Updated on 2024-07-18 at 13:53:33 +0000