def get_StaticMetropolis_instance(D, target_log_pdf):
step_size = 0.002
instance = StaticMetropolis(D, target_log_pdf, step_size)
# oracle scaling
instance.L_C = np.linalg.cholesky(true_cov)
return instance
def get_StaticMetropolis_instance(D, target_log_pdf):
step_size = 0.002
instance = StaticMetropolis(D, target_log_pdf, step_size)
# oracle scaling
instance.L_C = np.linalg.cholesky(true_cov)
return instance
def __init__(self, D, target_log_pdf, n, kernel_sigma, step_size, gamma2=0.1, schedule=None, acc_star=0.234):
StaticMetropolis.__init__(self, D, target_log_pdf, step_size, schedule, acc_star)
self.n = n
self.kernel_sigma = kernel_sigma
self.gamma2 = gamma2
self.Z = np.zeros((0, D))
def __init__(self, D, target_log_pdf, grad, step_size, schedule=None, acc_star=None):
StaticMetropolis.__init__(self, D, target_log_pdf, step_size, schedule, acc_star)
self.grad = grad
# members hidden from constructor
self.manual_gradient_step_size = None
self.do_preconditioning = False
self.forward_drift_norms = []
def get_StaticMetropolis_instance(D, target_log_pdf):
step_size = 0.002
acc_star = None
schedule = None
instance = StaticMetropolis(D, target_log_pdf, step_size, schedule, acc_star)
# give proposal variance a meaningful shape from previous samples
benchmark_samples_fname = "pmc_sv_benchmark_samples.txt"
benchmark_samples_sha1 = "d53e505730c41fbe413188530916d9a402e21a87"
assert_file_has_sha1sum(benchmark_samples_fname, benchmark_samples_sha1)
benchmark_samples = np.loadtxt(benchmark_samples_fname)
benchmark_samples = benchmark_samples[np.arange(0, len(benchmark_samples), step=50)]
instance.L_C = np.linalg.cholesky(np.cov(benchmark_samples.T))
return instance
def get_StaticMetropolis_instance(D, target_log_pdf):
step_size = 8.
schedule = one_over_sqrt_t_schedule
acc_star = 0.234
instance = StaticMetropolis(D, target_log_pdf, step_size, schedule, acc_star)
return instance
def __init__(self,
D,
target_log_pdf,
grad,
step_size,
schedule=None,
acc_star=None):
StaticMetropolis.__init__(self, D, target_log_pdf, step_size, schedule,
acc_star)
self.grad = grad
# members hidden from constructor
self.manual_gradient_step_size = None
self.do_preconditioning = False
self.forward_drift_norms = []
def __init__(self,
D,
target_log_pdf,
n,
kernel_sigma,
step_size,
gamma2=0.1,
schedule=None,
acc_star=0.234):
StaticMetropolis.__init__(self, D, target_log_pdf, step_size, schedule,
acc_star)
self.n = n
self.kernel_sigma = kernel_sigma
self.gamma2 = gamma2
self.Z = np.zeros((0, D))
def get_StaticMetropolis_instance(D, target_log_pdf):
step_size = 0.002
acc_star = None
schedule = None
instance = StaticMetropolis(D, target_log_pdf, step_size, schedule,
acc_star)
# give proposal variance a meaningful shape from previous samples
benchmark_samples_fname = "pmc_sv_benchmark_samples.txt"
benchmark_samples_sha1 = "d53e505730c41fbe413188530916d9a402e21a87"
assert_file_has_sha1sum(benchmark_samples_fname,
benchmark_samples_sha1)
benchmark_samples = np.loadtxt(benchmark_samples_fname)
benchmark_samples = benchmark_samples[np.arange(0,
len(benchmark_samples),
step=50)]
instance.L_C = np.linalg.cholesky(np.cov(benchmark_samples.T))
return instance
def get_StaticMetropolis_instance(D, target_log_pdf):
step_size = 1.
instance = StaticMetropolis(D, target_log_pdf, step_size)
return instance
