def ppm_group_algo_contour_plot_helper(test_name, fnames, algos,
granularity=config.PPM_CONTOUR_GRANULARITY,
alpha_start=config.PPM_ALPHA_START, alpha_end=config.PPM_ALPHA_END,
beta_start=config.PPM_BETA_START, beta_end=config.PPM_BETA_END):
if type(fnames) == str:
fnames = [fnames]
alpha_range = (float(alpha_start), float(alpha_end))
beta_range = (float(beta_start), float(beta_end))
granularity = int(granularity)
algos = algos.split(",")
kwargs = contour_settings(config.PPM_GROUP_CONTOUR_DEFAULT_ARGS,
config.PPM_GROUP_CONTOUR_OVERRIDES, test_name, fnames)
fig = plot_contour_base(xlim=beta_range, ylim=alpha_range)
colors = kwargs['colormap'](np.linspace(0, 1, len(algos)))
del kwargs['colormap']
for algo, color in zip(algos, colors):
prior, depth = algo.split("/")
depth = int(depth)
optimum, evals = benchmark.tasks.contour_data(prior, paranoia, depth,
alpha_range, beta_range, granularity, fnames)
label = short_name(config.SHORT_PRIOR_NAME, prior)
plot_optimum(optimum, label=label, color=color)
plot_contour_lines(optimum, evals, colors=color, **kwargs)
return plot.save_figure(fig, general.sanitize_fname(test_name), fnames)
def save_figure(fig, output_dir, fnames):
if len(fnames) == 1:
fig_fname = general.sanitize_fname(fnames[0])
else:
m = hashlib.md5()
for fname in fnames:
m.update(fname.encode('utf8'))
fig_fname = "group-" + m.hexdigest()
fig_fname += "_" + style_name + ".pdf"
fig_dir = os.path.join(config.FIGURE_DIR, output_dir)
os.makedirs(fig_dir, exist_ok=True)
fig_path = os.path.join(fig_dir, fig_fname)
with PdfPages(fig_path) as out:
print("Writing figure to " + fig_path)
out.savefig(fig)
return fig_path
def ppm_optimal_alpha_beta_helper(test_name, fname, prior,
depths=config.PPM_PARAMETER_DEPTHS,
granularity=config.PPM_PARAMETER_GRANULARITY):
depths = parse_depths(depths)
granularity = int(granularity)
csv_dir = os.path.join(config.TABLE_DIR, test_name)
os.makedirs(csv_dir, exist_ok=True)
csv_path = os.path.join(csv_dir, general.sanitize_fname(fname) + '.csv')
with open(csv_path, 'w') as f:
fieldnames = ['depth', 'alpha', 'beta', 'efficiency', 'status']
writer = csv.writer(f)
writer.writerow(fieldnames)
for d in depths:
res = benchmark.tasks.ppm_find_optimal_alpha_beta([fname], paranoia, prior, d, granularity,
config.PPM_ALPHA_RANGE, config.PPM_BETA_RANGE)
if res: # optimisation succeeded
(alpha, beta), efficiency, status = res
writer.writerow([d, alpha, beta, efficiency, status])
return csv_path