Rename the likelihood module to models (#35)

* rename likelihood module models * rename classes * update gwin executable * update samplers * update io module * update plot_movie * update unit tests * update init * update __init__ again * update option utils * change GaussianNoise.name to gaussian_noise * update test_option_utils * rename test_likelihood to test_models * fix whitespace * rename variable_args to model_params * update models * update samplers * update gwin executable * update io * update workflow utilities * update option_utils * update executables * update tests * update inference_test * update docs * rename static_args static_params * rename model_params variable_params * fix pep8 issues * more pep8
gwastro · Jul 13, 2018 · 7454fad · 7454fad
1 parent 67ee007
commit 7454fad
Show file tree

Hide file tree

Showing 27 changed files with 601 additions and 574 deletions.
diff --git a/bin/gwin b/bin/gwin
@@ -59,7 +59,7 @@ parser.add_argument("--seed", type=int, default=0,
 parser.add_argument("--samples-file", default=None,
                     help="Use an iteration from an InferenceFile as the "
                          "initial proposal distribution. The same "
-                         "number of walkers and the same [variable_args] "
+                         "number of walkers and the same [variable_params] "
                          "section in the configuration file should be used. "
                          "The priors must allow encompass the initial "
                          "positions from the InferenceFile being read.")
@@ -165,9 +165,6 @@ else:
                      "see --help")
 
 
-# get likelihood class
-likelihood_args = {}
-
 # set seed
 numpy.random.seed(opts.seed)
 logging.info("Using seed %i", opts.seed)
@@ -180,18 +177,18 @@ numpy.seterr(divide='ignore', invalid='ignore')
 ctx = scheme.from_cli(opts)
 fft.from_cli(opts)
 
-# Set up likelihood arguments
-likelihood_args = {}
+# Set up model arguments
+model_args = {}
 
 # get the data and psd
 strain_dict, stilde_dict, psd_dict = option_utils.data_from_cli(opts)
 low_frequency_cutoff_dict = option_utils.low_frequency_cutoff_from_cli(opts)
 if stilde_dict:
-    likelihood_args['data'] = stilde_dict
-    likelihood_args['f_lower'] = low_frequency_cutoff_dict.values()[0]
-    likelihood_args['delta_f'] = stilde_dict.values()[0].delta_f
-    likelihood_args['delta_t'] = strain_dict.values()[0].delta_t
-    likelihood_args['psds'] = psd_dict
+    model_args['data'] = stilde_dict
+    model_args['f_lower'] = low_frequency_cutoff_dict.values()[0]
+    model_args['delta_f'] = stilde_dict.values()[0].delta_f
+    model_args['delta_t'] = strain_dict.values()[0].delta_t
+    model_args['psds'] = psd_dict
 
 with ctx:
 
@@ -203,27 +200,26 @@ with ctx:
         logging.info("Initializing calibration model")
         recalib = {ifo : strain.read_model_from_config(cp, ifo) for
                    ifo in opts.instruments}
-        likelihood_args['recalib'] = recalib
+        model_args['recalib'] = recalib
 
     # get gates for templates
     gates = strain.gates_from_cli(opts)
     if gates:
-        likelihood_args['gates'] = gates
+        model_args['gates'] = gates
 
-    logging.info("Setting up likelihood")
+    logging.info("Setting up model")
 
     # construct class that will return the natural logarithm of likelihood
-    likelihood = gwin.likelihood.read_from_config(cp, section="likelihood",
-        **likelihood_args)
+    model = gwin.models.read_from_config(cp, section="model",
+        **model_args)
 
     burn_in_eval = burn_in.BurnIn(opts.burn_in_function,
                                 min_iterations=opts.min_burn_in)
 
-    # for likelihood evaluator
     logging.info("Setting up sampler")
 
     # create sampler that will run
-    sampler = option_utils.sampler_from_cli(opts, likelihood)
+    sampler = option_utils.sampler_from_cli(opts, model)
 
     # save information about this data and settings
     if not checkpoint_valid:
@@ -274,7 +270,9 @@ with ctx:
     elif len(cp.get_subsections("initial")):
         initial_dists = distributions.read_distributions_from_config(
             cp, section="initial")
-        init_prior = distributions.JointDistribution(variable_args,
+        constraints = distributions.read_constraints_from_config(cp,
+            constraint_section="initial_constraint")
+        init_prior = distributions.JointDistribution(sampler.variable_params,
             *initial_dists, **{"constraints" : constraints})
     else:
         init_prior = None
@@ -300,12 +298,12 @@ with ctx:
                 sampler.write_burn_in_iterations(fp, burnidx, is_burned_in)
                 # write the burn in results
                 logging.info("Writing burn in samples to file")
-                sampler.write_results(fp, static_args=likelihood.static_args,
+                sampler.write_results(fp, static_params=model.static_params,
                                       ifos=opts.instruments)
             # write to backup file
             with InferenceFile(backup_file, "a") as fp:
                 sampler.write_burn_in_iterations(fp, burnidx, is_burned_in)
-                sampler.write_results(fp, static_args=likelihood.static_args,
+                sampler.write_results(fp, static_params=model.static_params,
                                       ifos=opts.instruments)
 
 
@@ -356,7 +354,7 @@ with ctx:
         with InferenceFile(checkpoint_file, "a") as fp:
 
             logging.info("Writing results to file")
-            sampler.write_results(fp,static_args=likelihood.static_args,
+            sampler.write_results(fp, static_params=model.static_params,
                                   ifos=opts.instruments)
             logging.info("Updating burn in")
             burnidx, is_burned_in = burn_in_eval.update(sampler, fp)
@@ -373,7 +371,7 @@ with ctx:
         with InferenceFile(backup_file, "a") as fp:
 
             logging.info("Writing to backup file")
-            sampler.write_results(fp,static_args=likelihood.static_args,
+            sampler.write_results(fp, static_params=model.static_params,
                                   ifos=opts.instruments)
             sampler.write_burn_in_iterations(fp, burnidx, is_burned_in)
             if acls is not None:

diff --git a/bin/gwin_make_inj_workflow b/bin/gwin_make_inj_workflow
@@ -198,7 +198,7 @@ for i in range(n_injections):
     # make node for writing HTML table of parameters
     post_table_files += gwin_workflow.make_inference_summary_table(
                           workflow, inference_file, rdir["posteriors"],
-                          variable_args=unique_plot_parameters,
+                          variable_params=unique_plot_parameters,
                           analysis_seg=workflow.analysis_time,
                           tags=[str(i)])
 
@@ -256,12 +256,12 @@ if not opts.data_type == "analytical":
     inj_int_files = gwin_workflow.make_inference_inj_plots(
                                          workflow,
                                          inference_files, rdir.base,
-                                         cp.options("variable_args"),
+                                         cp.options("variable_params"),
                                          name="inference_intervals")
     inj_rec_files = gwin_workflow.make_inference_inj_plots(
                                          workflow,
                                          inference_files, rdir.base,
-                                         cp.options("variable_args"),
+                                         cp.options("variable_params"),
                                          name="inference_recovery")
     layout.two_column_layout(rdir.base,
                              [(a, b)

diff --git a/bin/gwin_make_workflow b/bin/gwin_make_workflow
@@ -320,7 +320,7 @@ for num_event in range(num_events):
     base = "posteriors"
     post_table_files = inffu.make_inference_summary_table(
                           workflow, inference_file, rdir[base],
-                          variable_args=all_parameters,
+                          variable_params=all_parameters,
                           analysis_seg=analysis_time,
                           tags=opts.tags + [str(num_event)])
     post_files = inffu.make_inference_posterior_plot(

diff --git a/bin/gwin_plot_movie b/bin/gwin_plot_movie
@@ -192,7 +192,7 @@ else:
 
 # get samples from InferenceFile
 file_parameters, trans = transforms.get_common_cbc_transforms(
-                                         parameters, fp.variable_args)
+                                         parameters, fp.variable_params)
 samples = fp.read_samples(file_parameters, thin_start=thin_start,
                           thin_interval=thinint, thin_end=thin_end,
                           iteration=itermask, flatten=False)
@@ -204,14 +204,14 @@ if samples.ndim > 2:
 
 # Get z-values
 if opts.z_arg is not None:
-    logging.info("Getting likelihood stats")
-    likelihood_stats = fp.read_likelihood_stats(thin_start=thin_start,
+    logging.info("Getting model stats")
+    model_stats = fp.read_model_stats(thin_start=thin_start,
         thin_end=thin_end, thin_interval=thinint, iteration=itermask,
         flatten=False)
-    if likelihood_stats.ndim > 2:
-        _, ii, jj = likelihood_stats.shape
-        likelihood_stats = likelihood_stats.reshape((ii, jj))
-    zvals, zlbl = option_utils.get_zvalues(fp, opts.z_arg, likelihood_stats)
+    if model_stats.ndim > 2:
+        _, ii, jj = model_stats.shape
+        model_stats = model_stats.reshape((ii, jj))
+    zvals, zlbl = option_utils.get_zvalues(fp, opts.z_arg, model_stats)
     show_colorbar = True
     # Set common min and max for colorbar in all plots
     if opts.vmin is None:

diff --git a/bin/gwin_plot_prior b/bin/gwin_plot_prior
@@ -84,7 +84,7 @@ cp = WorkflowConfigParser(opts.config_files)
 # if only section then look for subsections
 # and add distributions to list
 logging.info("Constructing prior")
-variable_args = []
+variable_params = []
 dists = []
 for sec in opts.sections:
     section = sec.split("-")[0]
@@ -97,21 +97,21 @@ for sec in opts.sections:
         name = cp.get_opt_tag(section, "name", subsection)
         dist = distributions.distribs[name].from_config(
                                             cp, section, subsection)
-        variable_args += dist.params
+        variable_params += dist.params
         dists.append(dist)
-variable_args = sorted(variable_args)
-ndim = len(variable_args)
+variable_params = sorted(variable_params)
+ndim = len(variable_params)
 
 # construct class that will return draws from the prior
-prior = distributions.JointDistribution(variable_args, *dists)
+prior = distributions.JointDistribution(variable_params, *dists)
 
 # get all points in space to calculate PDF
 logging.info("Getting grid of points")
 vals = numpy.zeros(shape=(ndim,opts.bins))
 bounds = [{}] * ndim
 for dist in dists:
     for param in dist.params:
-        idx = variable_args.index(param)
+        idx = variable_params.index(param)
         step = float(dist.bounds[param][1]-dist.bounds[param][0]) / opts.bins
         vals[idx,:] = numpy.arange(dist.bounds[param][0],dist.bounds[param][1],step)
         bounds[idx] = dist.bounds
@@ -121,13 +121,13 @@ pts = cartesian(vals)
 logging.info("Calculating PDF")
 pdf = []
 for pt in pts:
-    pt_dict = dict([(param,pt[j]) for j,param in enumerate(variable_args)])
+    pt_dict = dict([(param,pt[j]) for j,param in enumerate(variable_params)])
     pdf.append(sum([dist.pdf(**pt_dict) for dist in dists]))
 pdf = numpy.array(pdf)
 
 # check if only one parameter to plot PDF
 logging.info("Plotting") 
-if len(variable_args) == 1:
+if len(variable_params) == 1:
     x = vals[0,:]
     xmax = x.max()
     xmin = x.min()
@@ -136,12 +136,12 @@ if len(variable_args) == 1:
     plt.plot(x, pdf, "k", label="Prior")
     plt.xlim(xmin-pad, xmax+pad)
     plt.ylabel("Probability Density Function")
-    plt.xlabel(variable_args[0])
+    plt.xlabel(variable_params[0])
     plt.legend()
 
 # else make corner plot of all PDF
 else:
-    fig = corner.corner(pts, weights=pdf, labels=variable_args,
+    fig = corner.corner(pts, weights=pdf, labels=variable_params,
                        plot_contours=False, plot_datapoints=False)
 
     # remove the 1-D histograms
@@ -157,7 +157,7 @@ else:
 
 # save figure with meta-data
 caption_kwargs = {
-    "parameters" : ", ".join([param for param in variable_args])
+    "parameters" : ", ".join([param for param in variable_params])
 }
 caption = """This plot shows the probability density function (PDF) from the 
 prior distributions."""

diff --git a/bin/gwin_plot_samples b/bin/gwin_plot_samples
@@ -76,7 +76,7 @@ for i, arg in enumerate(parameters):
 
         # plot each walker as a different line on the subplot
         file_parameters, cs = transforms.get_common_cbc_transforms(
-                                                 parameters, fp.variable_args)
+                                                 parameters, fp.variable_params)
         y = fp.read_samples(file_parameters, walkers=j,
                             thin_start=opts.thin_start,
                             thin_interval=opts.thin_interval,