import numpy as np
import matplotlib.pyplot as plt
import mcmc
# Potential for a 1dim Gaussian
def U(x):
return x**2 / 2
def gradU(x):
return x
h_metropolis = 5
num_samples = 50000
skip_n_samples = 5
#Ignore the further results of the chain, useful only for convergence analysis
X, runtime, _, _ = mcmc.chain_rwMetropolis(np.random.random_sample(1),
h_metropolis, U, num_samples, skip_n_samples)
info_str = "INFOSIMU: chain_rwMetropolis, h = " + str(h_metropolis) + \
" runtime: " + runtime + " n_samples = " + str(num_samples) + '\n'
# Store the samples into a separate file, to incentivate a C approach
filename = "gaussian_stored_samples.smp"
samples_file = open(filename, "w")
samples_file.write(info_str)
for x in X:
for i in x:
print(i, file = samples_file)
samples_file.close()
print("Samples and information stored in " + filename)
MALA = False
MULTICHAIN_RW = True
# 10 chains to produce in parallel
multich = mp.cpu_count()
#multich = 10
dim = 23
if SAMPLING_SINGLE_CHAIN:
print("Constructing a single full chain")
if METROPOLIS_RW:
print("...Metropolis RW")
# X, runtime, _, _ = \
# mcmc.chain_rwMetropolis(np.random.random_sample(23),
# h, auxiliary_tot_cost, num_samples, skip_n_samples, L_domain)
X, runtime, _, _ = mcmc.chain_rwMetropolis(np.random.random_sample(23),
h, U, num_samples,
skip_n_samples, L_domain)
info_str = "INFOSIMU: chain_rwMetropolis, h = " + str(h) + \
" runtime: " + runtime + " n_samples = " + str(num_samples) + '\n'
elif ULA:
print("...ULA")
X, runtime, _ = mcmc.ulaChain(np.random.random_sample(23), h, U, gradU,
num_samples, skip_n_samples, L_domain)
info_str = "INFOSIMU: ULA, h = " + str(h) + \
" runtime: " + runtime + " n_samples = " + str(num_samples)+'\n'
elif MALA:
print("...MALA")
X, runtime, _, _ = mcmc.malaChain(np.random.random_sample(23), h, U,
gradU, num_samples, skip_n_samples,
L_domain)
info_str = "INFOSIMU: MALA, h = " + str(h) + \
METROPOLIS_RW = False #True
ULA = False #True
MALA = False
MULTICHAIN_RW = True #False #True
# 10 chains to produce in parallel
multich = mp.cpu_count()
#multich = 10
dim = 401
if SAMPLING_SINGLE_CHAIN:
print("Constructing a single full chain")
if METROPOLIS_RW:
print("...Metropolis RW")
X, runtime, _, _ = \
mcmc.chain_rwMetropolis(np.random.random_sample(dim),
h, auxiliary_tot_cost, num_samples, skip_n_samples, L_domain)
info_str = "INFOSIMU: chain_rwMetropolis, h = " + str(h) + \
" runtime: " + runtime + " n_samples = " + str(num_samples) + '\n'
elif MULTICHAIN_RW:
print("...multichain RW approach")
X, arate, _ = \
mcmc.multichainRW(dim, L_domain, h, U, num_samples, multich,
skip_n_samples, True)
info_str = "INFOSIMU: Multichain RW"
# Store the samples into a separate file, modular approach
filename = "regression_chain.smp"
if (len(sys.argv) == 2):
filename = str(sys.argv[1]) + "_chain.smp"
samples_file = open(filename, "w")
W2, W3, W4, 10) * k_enlarge
# The gradient of the potential above
def auxiliary_gradient(params_together):
b2, b3, b4, W2, W3, W4 = nn_potential.split_params(params_together)
return nn_potential.grad_cost(nn_potential.x, nn_potential.y, b2, b3, b4,
W2, W3, W4) * k_enlarge
# Sample candidated from HMC to minimize the potential
#X, runtime = mcmc.sample_uHMC(np.random.random_sample(23) * 0.5,
# 0.005, 0.0005, 0.5, auxiliary_gradient, 500000)
X, runtime, _, _ = mcmc.chain_rwMetropolis(
np.random.random_sample(23) * 0.5, h_metropolis, auxiliary_tot_cost,
num_samples, skip_n_samples, L_domain)
info_str = "INFOSIMU: chain_rwMetropolis, h = " + str(h_metropolis) + \
" runtime: " + runtime + " n_samples = " + str(num_samples) + '\n'
# Store the samples into a separate file, to incentivate a chain approach
filename = "nn_stored_samples.smp"
if (len(sys.argv) == 2):
filename = str(sys.argv[1]) + ".smp"
samples_file = open(filename, "w")
samples_file.write(info_str)
for x in X:
for i in range(len(x)):
if i < (len(x) - 1):
print(x[i], file=samples_file, end=' ')
else:
METROPOLIS_RW = False #True
ULA = False
MALA = False #True
MULTICHAIN_RW = True
# 10 chains to produce in parallel
#multich = 20
multich = mp.cpu_count()
dim = 2
if SAMPLING_SINGLE_CHAIN:
print("Sampling from a single chain")
if METROPOLIS_RW:
print("Metropolis RW")
X, runtime, _, _ = mcmc.chain_rwMetropolis(np.array([4, 1]), h, ban_U,
num_samples, skip_n_samples,
L_domain)
info_str = "INFOSIMU: chain_rwMetropolis, h = " + str(h) + \
" runtime: " + runtime + " n_samples = " + str(num_samples)+'\n'
elif ULA:
print("ULA")
X, runtime, _ = mcmc.ulaChain(np.array([4, 1]), h, ban_U, ban_gradU,
num_samples, skip_n_samples, L_domain)
info_str = "INFOSIMU: ULA, h = " + str(h) + \
" runtime: " + runtime + " n_samples = " + str(num_samples)+'\n'
elif MALA:
print("MALA")
X, runtime, _, _ = mcmc.malaChain(np.array([4, 1]), h, ban_U,
ban_gradU, num_samples,
skip_n_samples, L_domain)
info_str = "INFOSIMU: MALA, h = " + str(h) + \
