py_sgd_params_ps = pyGradientDescentParameters(
max_num_steps=1000,
max_num_restarts=3,
num_steps_averaged=15,
gamma=0.7,
pre_mult=1.0,
max_relative_change=0.02,
tolerance=1.0e-10,
)
cpp_sgd_params_ps = cppGradientDescentParameters(
num_multistarts=1,
max_num_steps=6,
max_num_restarts=1,
num_steps_averaged=3,
gamma=0.0,
pre_mult=1.0,
max_relative_change=0.1,
tolerance=1.0e-10,
)
cpp_sgd_params_kg = cppGradientDescentParameters(
num_multistarts=200,
max_num_steps=50,
max_num_restarts=2,
num_steps_averaged=4,
gamma=0.7,
pre_mult=1.0,
max_relative_change=0.5,
tolerance=1.0e-10,
)
n_hypers=10,
noisy=False)
cpp_gp_loglikelihood.train()
py_sgd_params_ps = pyGradientDescentParameters(max_num_steps=200,
max_num_restarts=2,
num_steps_averaged=15,
gamma=0.7,
pre_mult=0.01,
max_relative_change=0.1,
tolerance=1.0e-5)
cpp_sgd_params_ps = cppGradientDescentParameters(num_multistarts=1,
max_num_steps=20,
max_num_restarts=2,
num_steps_averaged=3,
gamma=0.7,
pre_mult=0.03,
max_relative_change=0.06,
tolerance=1.0e-5)
cpp_sgd_params_kg = cppGradientDescentParameters(num_multistarts=200,
max_num_steps=30,
max_num_restarts=2,
num_steps_averaged=4,
gamma=0.7,
pre_mult=0.3,
max_relative_change=0.3,
tolerance=1.0e-5)
# minimum of the mean surface
eval_pts = inner_search_domain.generate_uniform_random_points_in_domain(
max_relative_change=0.02,
tolerance=1.0e-8)
py_sgd_params_acquisition = pyGradientDescentParameters(
max_num_steps=50,
max_num_restarts=1,
num_steps_averaged=0,
gamma=0.7,
pre_mult=1.0,
max_relative_change=0.1,
tolerance=1.0e-8)
cpp_sgd_params_ps = cppGradientDescentParameters(num_multistarts=1,
max_num_steps=6,
max_num_restarts=1,
num_steps_averaged=3,
gamma=0.0,
pre_mult=1.0,
max_relative_change=0.1,
tolerance=1.0e-8)
cpp_sgd_params_kg = cppGradientDescentParameters(num_multistarts=int(1000),
max_num_steps=60,
max_num_restarts=3,
num_steps_averaged=0,
gamma=0.7,
pre_mult=1.0,
max_relative_change=0.1,
tolerance=1.0e-8)
# minimum of the mean surface
cpp_gp = cpp_gp_loglikelihood.models[0]
# constants
obj_func_dict = {
'Branin': synthetic_functions.Branin(),
'Hartmann': synthetic_functions.Hartmann(),
'Rosenbrock': synthetic_functions.Rosenbrock(),
'Ackley': synthetic_functions.Ackley(),
'Levy': synthetic_functions.Levy()
}
# opt_method = {'qKGg': bgo_methods.gen_sample_from_qkg(), 'qEIg': bgo_methods.gen_sample_from_qei()}
cpp_sgd_params_ei = cppGradientDescentParameters(num_multistarts=200,
max_num_steps=100,
max_num_restarts=2,
num_steps_averaged=15,
gamma=0.7,
pre_mult=0.1,
max_relative_change=0.7,
tolerance=1.0e-5)
cpp_sgd_params_hyper = cppGradientDescentParameters(num_multistarts=100,
max_num_steps=100,
max_num_restarts=2,
num_steps_averaged=15,
gamma=0.7,
pre_mult=0.1,
max_relative_change=0.02,
tolerance=1.0e-5)
cpp_sgd_params_kg = cppGradientDescentParameters(num_multistarts=50,
max_num_steps=100,
# constants
a = numpy.random.normal(0, 1)
theta = 10 * numpy.random.multivariate_normal(numpy.zeros(6),
numpy.identity(6))
obj_func_dict = {
'Branin': synthetic_functions.Branin(),
'Hartmann': synthetic_functions.Hartmann3(),
'Rosenbrock': synthetic_functions.Rosenbrock(),
'Ackley': synthetic_functions.Ackley(),
'Levy': synthetic_functions.Levy()
}
cpp_sgd_params = cppGradientDescentParameters(num_multistarts=2000,
max_num_steps=20,
max_num_restarts=1,
num_steps_averaged=15,
gamma=0.7,
pre_mult=1.0,
max_relative_change=0.7,
tolerance=1.0e-3)
cpp_sgd_params_ps = cppGradientDescentParameters(num_multistarts=1,
max_num_steps=12,
max_num_restarts=1,
num_steps_averaged=3,
gamma=0.7,
pre_mult=0.01,
max_relative_change=0.01,
tolerance=1.0e-5)
if obj_func_name == "GP":
gp_grad_info_dict = pickle.load(open('random_gp_grad_1d', 'rb'))
