def __init__(self, num_dimensions, num_initial_components,
initial_mixture_prior_variance, cart_likelihood_var,
conf_likelihood_var, config):
self.cart_likelihood_var = cart_likelihood_var
self.conf_likelihood_var = conf_likelihood_var
filepath = "planar_n_link/" + str(
num_dimensions) + '/' + datetime.datetime.now().strftime(
"%Y-%m-%d_%H%M%S") + '/'
AbstractVIPSExperiment.__init__(self, filepath, num_dimensions,
num_initial_components,
initial_mixture_prior_variance)
[self.target_lnpdf, self.prior, self.prior_chol
] = build_target_likelihood_planar_n_link(num_dimensions,
conf_likelihood_var,
cart_likelihood_var)
self.groundtruth_samples = None
self.groundtruth_lnpdfs = None
self.config = config
# add plot for robot visualization
def all_plots(sampler):
plt.figure(100)
plt.clf()
[weights, means, _] = sampler.vips_c.get_model()
visualize_mixture(weights, means)
default_plots(sampler)
changes_in_functions = {
'plot_fctn': all_plots,
}
ConfigUtils.merge_FUNCTIONS(config, changes_in_functions)
import numpy as np
from sampler.PTMCMC.PTMCMCSampler.PTMCMCSampler import PTSampler
from experiments.lnpdfs.create_target_lnpfs import build_target_likelihood_planar_n_link
from time import time
from mpi4py import MPI
num_dimensions = 10
conf_likelihood_var = 4e-2 * np.ones(num_dimensions)
conf_likelihood_var[0] = 1
cart_likelihood_var = np.array([1e-4, 1e-4])
[target_lnpdf, prior,
prior_chol] = build_target_likelihood_planar_n_link(num_dimensions,
conf_likelihood_var,
cart_likelihood_var)
def sample(n, path, known_prior, comm=MPI.COMM_WORLD):
if known_prior:
def lnlikefn(theta):
lnlikefn.counter += 1
return target_lnpdf(theta, without_prior=True)
def lnpriorfn(theta):
return prior.logpdf(theta)
else:
def lnlikefn(theta):
lnlikefn.counter += 1
return target_lnpdf(theta, without_prior=False)
from scipy.stats import multivariate_normal as normal
import numpy as np
from time import time
from experiments.lnpdfs.create_target_lnpfs import build_target_likelihood_planar_n_link
from sampler.SliceSampling.slice_sampler import slice_sample
num_dimensions = 3
conf_likelihood_var = 4e-2 * np.ones(num_dimensions)
conf_likelihood_var[0] = 1
cart_likelihood_var = np.array([1e-4, 1e-4])
lnpdf = build_target_likelihood_planar_n_link(num_dimensions,
conf_likelihood_var,
cart_likelihood_var)[0]
prior = normal(np.zeros((num_dimensions)),
conf_likelihood_var * np.eye((num_dimensions)))
initial = prior.rvs(1)
def sample(n_samps, sigma, path):
start = time()
[samples, fevals,
timestamps] = slice_sample(lnpdf, initial, n_samps,
sigma * np.ones(num_dimensions))
timestamps -= start
samples = samples.transpose().reshape(len(timestamps), -1, 3).copy()
np.savez(path + 'processed_data',
samples=samples,
fevals=fevals,
timestamps=timestamps)
