np.savetxt(input_data_file, np.transpose([xin, yin]))
## Run the inference code #####################################################
ii = 0
for idim in chdimlist:
for line in fileinput.input("line_infer.xml", inplace=1):
print(line.replace('PAR_' + str(idim), str(chstart[ii])), end=" ")
ii = ii + 1
print(ii)
print("Running the parameter inference")
os.system('./line_infer.x >& logMCMC.dat')
## Import data from MCMC file ###################################################
print("Loading in chain file ", chainfile)
all_samples, vnames = postp.extract_all_vars(chainfile, n_burnin, 0, stride)
n_all_vars = len(vnames)
print(all_samples.shape)
# Extract all MCMC chain variables in separate array
chn = all_samples[:, 0:1 + n_all_vars]
nchn = chn.shape[0]
if generate_plots:
## Find posterior predictive #######################################################
# ideally we should have a deterministic forward model code,
# but for a line it is simple enough to compute here
print("\nSampling chain to compute posterior predictive distribution\n")
# First set parameters to the default parameters; then replace the inferred parameters
# with samples from the chain.
stride = model.model_info["stride"]
if demo_verbose > 1:
print("n_skip:", n_skip)
print("stride:", stride)
# Base name of file for outputting results
samples_file_name = model.model_info["output_file"] #MCMC file name
out_file_base = samples_file_name + ".sk" + str(n_skip) + ".st" + str(
stride)
# Import variables of interest from the MCMC data file
print("UQTk extract all vars")
all_samples, v_names = uqtkinfpp.extract_all_vars(
samples_file_name,
n_skip,
demo_verbose,
stride,
labels=labels_present)
# Get statistics
print("UQTk get_mcmc_stats")
map_params = uqtkinfpp.get_mcmc_stats(all_samples, v_names,
out_file_base, demo_verbose,
desired_stats)
print("\nMAP parameter set:")
for idx in range(len(map_params)):
print(v_names[idx], ":", map_params[idx])
# Save MAP parameter info to a file
MAP_info = np.stack((v_names, map_params), axis=-1)
generate_plots = False
else:
assert False, "unhandled command line parsing option"
# Other input settings
dense_plots = True # Set to True to get dense format for posterior plots in triangular format
chainfile = 'tmcmc_chain.dat' # Name of the chain file
## Run the inference code #####################################################
print("Running TMCMC on bimodal posterior")
os.system('./tmcmc_bimodal.x >& logTMCMC.dat')
## Import data from MCMC file ###################################################
print("Loading in chain file ", chainfile)
all_samples, vnames = postp.extract_all_vars(chainfile, 0, 0, 1, False)
n_all_vars = len(vnames)
print(all_samples.shape)
# Extract all MCMC chain variables in separate array
chn = all_samples[:, 0:1 + n_all_vars]
nchn = chn.shape[0]
if generate_plots:
## Scatter plots from intermediate pdfs ##################################################################
indx = 0
while 1:
fname = './TMCMCIntermediates/samples.dat.' + str(indx)
if os.path.exists(fname):