file src/ObjectivesBit.cpp
[No description available] More…
Namespaces
| Name |
|---|
| Gambit TODO: see if we can use this one: |
| Gambit::ObjectivesBit |
Detailed Description
Author: Andrew Fowlie (andrew.j.fowlie@qq.com)
Date: 2021 January
Functions of module ObjectivesBit
Put your functions in files like this if you wish to add observables or likelihoods to this module.
See https://en.wikipedia.org/wiki/Test_functions_for_optimizationhttps://arxiv.org/abs/2101.04525 and https://arxiv.org/abs/1306.2144
Authors (add name and date if you modify):
Source code
// GAMBIT: Global and Modular BSM Inference Tool
// *********************************************
/// \file
///
/// Functions of module ObjectivesBit
///
/// Put your functions in files like this
/// if you wish to add observables or likelihoods
/// to this module.
///
/// See https://en.wikipedia.org/wiki/Test_functions_for_optimization
/// https://arxiv.org/abs/2101.04525
/// and https://arxiv.org/abs/1306.2144
///
/// *********************************************
///
/// Authors (add name and date if you modify):
///
/// \author Andrew Fowlie
/// (andrew.j.fowlie@qq.com)
/// \date 2021 January
///
/// *********************************************
#include <cmath>
#include <limits>
#include <vector>
#include "gambit/Elements/gambit_module_headers.hpp"
#include "gambit/ObjectivesBit/ObjectivesBit_rollcall.hpp"
namespace Gambit
{
namespace ObjectivesBit
{
typedef Gambit::Models::safe_param_map<Gambit::safe_ptr<const double>> map;
std::vector<double> get_arguments(map param)
{
std::vector<double> x;
for (int i = 0, n = param.size(); i < n; i++)
{
std::string name = "x" + std::to_string(i + 1);
x.push_back(*param.at(name));
}
return x;
}
void gaussian(double &loglike)
{
using namespace Pipes::gaussian;
auto x = get_arguments(Param);
const double mu = 0.5;
const double sigma = 0.1;
double chi_sq = 0.;
for (const auto& e : x)
{
chi_sq += (e - mu) * (e - mu);
}
chi_sq /= sigma * sigma;
static double norm = 0.5 * x.size() * std::log(2. * M_PI * sigma * sigma);
loglike = -0.5 * chi_sq - norm;
}
double rosenbrock(double x, double y)
{
const double a = 1.;
const double b = 100.;
return (a - x) * (a - x) + b * (y - x * x) * (y - x * x);
}
double rosenbrock(std::vector<double> x)
{
const int d = x.size();
double r = 0.;
for (int i = 0; i < d - 1; i++)
{
r += rosenbrock(x[i], x[i + 1]);
}
return r;
}
/** @brief See https://en.wikipedia.org/wiki/Rosenbrock_function */
void rosenbrock(double &loglike)
{
using namespace Pipes::rosenbrock;
auto x = get_arguments(Param);
loglike = - rosenbrock(x);
}
/** @brief See https://en.wikipedia.org/wiki/Himmelblau%27s_function */
void himmelblau(double &loglike)
{
using namespace Pipes::himmelblau;
auto x = get_arguments(Param);
loglike = - (x[0] * x[0] + x[1] - 11.) * (x[0] * x[0] + x[1] - 11.)
- (x[0] + x[1] * x[1] - 7.) * (x[0] + x[1] * x[1] - 7.);
}
void mccormick(double &loglike)
{
using namespace Pipes::mccormick;
auto x = get_arguments(Param);
loglike = - std::sin(x[0] + x[1])
- (x[0] - x[1]) * (x[0] - x[1])
+ 1.5 * x[0] - 2.5 * x[1] - 1.;
}
/** @brief See https://en.wikipedia.org/wiki/Ackley_function */
void ackley(double &loglike)
{
using namespace Pipes::ackley;
auto x = get_arguments(Param);
double r2 = 0.;
double c = 0.;
for (const auto& p : x)
{
r2 += p * p;
c += std::cos(2. * M_PI * p);
}
loglike = 20. * std::exp(-0.2 * std::sqrt(r2 / x.size()))
+ std::exp(c / x.size())
- M_E - 20.;
}
/** @brief Test problem 2 from https://arxiv.org/abs/1306.2144 */
void eggbox(double &loglike)
{
using namespace Pipes::eggbox;
auto x = get_arguments(Param);
double prod = 1.;
for (const auto& p : x)
{
prod *= std::cos(p * 5. * M_PI);
}
loglike = std::pow(prod + 2., 5.);
}
/** @brief See https://en.wikipedia.org/wiki/Rastrigin_function */
void rastrigin(double &loglike)
{
using namespace Pipes::rastrigin;
auto x = get_arguments(Param);
double r2 = 0.;
double c = 0.;
for (const auto& p : x)
{
r2 += p * p;
c += std::cos(2. * M_PI * p);
}
loglike = x.size() * 10. + r2 - 10. * c;
}
void beale(double &loglike)
{
using namespace Pipes::beale;
auto x = get_arguments(Param);
loglike = - std::pow(1.5 - x[0] - x[1] * x[0], 2)
- std::pow(2.25 - x[0] - x[1] * x[1] * x[0], 2)
- std::pow(2.625 - x[0] - x[1] * x[1] * x[1] * x[0], 2);
}
double logaddexp(double x, double y)
{
return std::max(x, y) + std::log1p(std::exp(-std::abs(x - y)));
}
/** @brief Test problem 1 from https://arxiv.org/abs/1306.2144 */
void shells(double &loglike)
{
using namespace Pipes::shells;
auto x = get_arguments(Param);
const double radius = 0.1;
const double width = 2.;
const std::vector<double> center = {-3.5, 3.5};
static double norm = - 0.5 * std::log(2. * M_PI * width * width);
loglike = std::numeric_limits<double>::lowest();
for (const auto& c : center)
{
double r2 = 0.;
for (const auto& p : x)
{
r2 += (c - p) * (c - p);
}
double d = (std::sqrt(r2) - radius) * (std::sqrt(r2) - radius);
double delta = -d / (2. * width * width) - norm;
loglike = logaddexp(loglike, delta);
}
}
void styblinski_tang(double &loglike)
{
using namespace Pipes::styblinski_tang;
auto x = get_arguments(Param);
loglike = 0.;
for (const auto& p : x)
{
double p2 = p * p;
loglike -= 0.5 * (p2 * p2 - 16. * p2 + 5. * p);
}
}
void easom(double &loglike)
{
using namespace Pipes::easom;
auto x = get_arguments(Param);
loglike = std::cos(x[0]) * std::cos(x[1]) *
std::exp(- (x[0] - M_PI) * (x[0] - M_PI) - (x[1] - M_PI) * (x[1] - M_PI));
}
/** @brief Analytic function 1 from https://arxiv.org/abs/2101.04525 */
void tf1(double &loglike)
{
using namespace Pipes::tf1;
auto x = get_arguments(Param);
const double location = 2.;
static const double scale = std::pow(15., 6);
double sum_x2 = 0.;
double sum_x6 = 0.;
double prod_c2 = 1.;
for (auto p : x)
{
p -= location;
double x2 = p * p;
sum_x2 += x2;
sum_x6 += x2 * x2 * x2;
double c = std::cos(p);
prod_c2 *= c * c;
}
loglike = - std::exp(-sum_x6 / scale) + 2. * std::exp(-sum_x2) * prod_c2;
}
/** @brief Analytic function 2 from https://arxiv.org/abs/2101.04525 */
void tf2(double &loglike)
{
using namespace Pipes::tf2;
auto x = get_arguments(Param);
const double location = -0.23;
loglike = 0.;
for (auto p : x)
{
p -= location;
loglike -= p * p - 10. * std::cos(2. * M_PI * p) + 10.;
}
}
/** @brief Analytic function 3 from https://arxiv.org/abs/2101.04525 */
void tf3(double &loglike)
{
using namespace Pipes::tf3;
auto x = get_arguments(Param);
const int n = x.size();
double sum_s6 = 0.;
for (const auto& p : x)
{
double s = std::sin(5. * M_PI * (std::pow(p, 0.75) - 0.05));
sum_s6 += std::pow(s, 6);
}
loglike = 1. / n * sum_s6;
}
/** @brief Analytic function 4 from https://arxiv.org/abs/2101.04525 */
void tf4(double &loglike)
{
using namespace Pipes::tf4;
auto x = get_arguments(Param);
const int n = x.size();
double sum = 0.;
for (const auto& p : x)
{
sum += p * std::sin(std::sqrt(std::abs(p)));
}
loglike = -418.982887 * n + sum;
}
} // namespace ObjectivesBit
} // namespace Gambit
Updated on 2023-06-26 at 21:36:54 +0000