kernel_width=nu)
beta_emp_store = np.zeros((n_exp,dim+1))
for i_exp in range(n_exp):
if np.mod(i_exp + 1,10) == 0:
s_exp = "Experiment {} / {} is running...".format(i_exp + 1,n_exp)
print(s_exp)
# getting the explanation
exp = explainer.explain_instance(xi,
my_model,
num_samples=n_samples)
# getting the coefficients of the local model
beta_emp_store[i_exp,:] = format_coefs(exp)
# computing the theoretical values
beta_theo = compute_beta_indicator(xi,rect,nu,my_stats)
###########################################################################
# plotting the results
fig, ax = plt.subplots(figsize=(15,10))
plot_whisker_boxes(beta_emp_store,
ax,
theo=beta_theo)
# saving the fig
s_name = "results/indicator_explanation"
fig.savefig(s_name + '.pdf',format='pdf',bbox_inches = 'tight',pad_inches = 0)
# hide_color = None gives mean replacement
hide_color = 0
rgb_hide_color = hide_color_helper(hide_color)
# entering the main loop
for i in tqdm(range(0,n_exp)):
# explanation
explanation = explainer.explain_instance(xi,
classifier_fn=basic_shape_detector,
top_labels=10,
hide_color=rgb_hide_color,
num_samples=n_examples,
segmentation_fn=segmenter)
# get the explanations
betahat_store[i,:] = format_coefs(explanation,0)
print("done!")
print()
# compute the theoretical explanations
exp_theo = compute_beta_basic_shape(xi,segments,indices,tau,hide_color=hide_color)
####################################################################
# this folder contains the resulting figure
fig_dir = "results/figures/"
mkdir(fig_dir)
# graphical parameters
plt.rcParams['pdf.fonttype'] = 42
s_exp = "Experiment {} / {} is running...".format(i_exp + 1, n_exp)
print(s_exp)
# getting the explanations for both models and the sum
exp_1 = explainer.explain_instance(xi,
my_model_1,
num_samples=n_samples)
exp_2 = explainer.explain_instance(xi,
my_model_2,
num_samples=n_samples)
exp_sum = explainer.explain_instance(xi,
model_sum,
num_samples=n_samples)
# getting the coefficients of the local models
beta_emp_store_1[i_exp, :] = format_coefs(exp_1)
beta_emp_store_2[i_exp, :] = format_coefs(exp_2)
beta_emp_store_sum[i_exp, :] = format_coefs(exp_sum)
beta_mean_1 = np.mean(beta_emp_store_1, 0)
beta_mean_2 = np.mean(beta_emp_store_2, 0)
beta_mean_sum = np.mean(beta_emp_store_sum, 0)
###########################################################################
# plotting the interpretable coefficients for the three models
fig_1, ax_1 = plt.subplots(figsize=(15, 10))
plot_whisker_boxes(beta_emp_store_1, ax_1, theo=beta_mean_1, color="k")
fig_2, ax_2 = plt.subplots(figsize=(15, 10))
plot_whisker_boxes(beta_emp_store_2, ax_2, theo=beta_mean_2, color="k")
predictions = model.predict(tf.expand_dims(xi_norm, 0))
predicted_class = np.argmax(predictions)
# get the segments
segments = segmenter(xi_norm)
d = np.unique(segments).shape[0]
# get the explanations
beta_store = np.zeros((n_exp, d + 1))
for i_exp in tqdm(range(n_exp)):
print("run {} / {}".format(i_exp + 1, n_exp))
explanation = explainer.explain_instance(xi_norm,
classifier_fn=model.predict,
num_samples=n_examples,
segmentation_fn=segmenter)
beta_store[i_exp, :] = format_coefs(explanation, predicted_class)
# store everything
empirical = {}
empirical["predictions"] = predictions
empirical["segments"] = segments
empirical["predicted_class"] = predicted_class
empirical["confidence"] = predictions[0, predicted_class]
empirical["explanations"] = beta_store
empirical["image_name"] = id_image + 1
pickle_name = result_path + str(id_image + 1).zfill(8) + '.pkl'
print("saving results...")
with open(pickle_name, 'wb') as f:
pickle.dump(empirical, f)
# hide_color = None gives mean replacement
hide_color = 0
rgb_hide_color = hide_color_helper(hide_color)
# entering the main loop
for i in tqdm(range(0, n_exp)):
# explanation
explanation = explainer.explain_instance(xi_rgb,
classifier_fn=linear_function,
top_labels=10,
hide_color=rgb_hide_color,
num_samples=n_examples,
segmentation_fn=segmenter)
data_store[i, :] = format_coefs(explanation, 0)
print("done!")
print()
# computing theory
exp_theo = compute_beta_linear(xi_vec, segments, coefs, hide_color=hide_color)
####################################################################
fig_dir = "results/figures/"
mkdir(fig_dir)
# plot the examples and superpixels
fig, axis = plt.subplots(1, 3, figsize=(16, 5))