linewidth=linewidth)
plt.title(method_name, fontsize=fontsize)
legend = plt.legend(LEGEND_NAMES,
fontsize=fontsize,
bbox_to_anchor=(1, 1),
loc=2,
borderaxespad=0.)
plt.xlim([0, 100])
plt.ylim([0.5, 1])
plt.xlabel('Number of experiments', fontsize=fontsize)
plt.ylabel('Prediction accuracy', fontsize=fontsize)
plt.tight_layout()
plot_filename = os.path.join(PLOT_FOLDER,
'learning_curve_{}'.format(method_name))
save_and_close_figure(fig, plot_filename, exts=['.png'], legend=legend)
##
class_names = ['NoCrystal', 'Crystal']
for iclass, class_name in enumerate(class_names):
fig = plt.figure(figsize=(12, 8))
for xp_name in XP_NAMES:
result_dict = RESULTS[xp_name]
plt.plot(result_dict['test_range'],
result_dict['class_acc'][iclass],
linewidth=linewidth)
plt.title('{} | {}'.format(class_name, method_name),
fontsize=fontsize)
legend = plt.legend(LEGEND_NAMES,
fontsize=fontsize,
AVG_RESULTS['RANDOM'],
'b',
linestyle='-',
linewidth=linewidth,
marker='D',
markersize=markersize)
plt.title('Evolution of Crystallization Model Quality',
fontsize=fontsize)
plt.legend(['Algorithm', 'Human', 'Random'], fontsize=fontsize, loc=2)
plt.xlim([-1, 101])
if method_name == 'Adaboost':
plt.ylim([60, 85])
y_tick_pos = [60, 65, 70, 75, 80, 85]
else:
plt.ylim([65, 85])
y_tick_pos = [65, 70, 75, 80, 85]
plt.yticks(y_tick_pos, ['{}%'.format(i) for i in y_tick_pos])
plt.xlabel('Number of Experiments', fontsize=fontsize)
plt.ylabel('Prediction Accuracy - %', fontsize=fontsize)
#
PLOT_FOLDER = os.path.join(HERE_PATH, 'plot')
filetools.ensure_dir(PLOT_FOLDER)
plot_filename = os.path.join(
PLOT_FOLDER, 'learning_curve_unbiased_{}'.format(method_name))
save_and_close_figure(fig, plot_filename)
fig = plt.figure(figsize=(12, 8))
all_data = []
for i, method_name in enumerate(method_names):
data = results[problem_name][method_name]['accuracy_score']
data = np.array(data)
if method_name == 'uncertainty_single':
data = data[:, 0:-1:10]
all_data.append(data)
all_data = np.array(all_data)
all_data = np.swapaxes(all_data, 0, 1)
all_data = np.swapaxes(all_data, 1, 2)
ax = sns.tsplot(all_data,
time=range(1, 11),
condition=method_names,
ci=[68, 95])
ax.plot([1, 10], [1, 1], color='grey', linestyle='--')
ax.set_xlabel('Iterations', fontsize=fontsize)
ax.set_ylabel('Prediction Accuracy', fontsize=fontsize)
ylim = ax.get_ylim()
ax.set_ylim([ylim[0], 1 + 0.05 * np.diff(ylim)])
ax.legend(bbox_to_anchor=(1, 0.35), fontsize=fontsize)
figures.append(fig)
#
for i, fig in enumerate(figures):
foldername = os.path.join(HERE_PATH, 'plot', problem_names[i])
filename = os.path.join(foldername, 'accuracy')
save_and_close_figure(fig, filename, exts=['.png'])