model = train(train_val_x, train_val_y, best_C)
t3 = time.time()
print("training time:", t3 - t2)
pred = predict(test_x, test_y, model)
t4 = time.time()
print("testing time:", t4 - t3)
accuracies.append(accuracy_score(pred, test_y) * 100)
print(accuracy_score(pred, test_y) * 100)
train_times.append(t3 - t2)
test_times.append(t4 - t3)
mean_accuracies.append(mean(accuracies))
acc_list.append(mean(mean_accuracies))
std_list.append(stdev(mean_accuracies))
train_time_list.append(mean(train_times))
test_time_list.append(mean(test_times))
dict_to_csv(
{
'accuracy': acc_list,
'error': std_list,
'train_time': train_time_list,
'test_time': test_time_list,
'landmarks': landmarks
}, ["nb_iter={},cv={}".format(ITER, 3)], PATH + ".csv")
7), labels[train_val_inds]
models = train_svm_per_view(train_val_x, train_val_y, 17, 7, best_C)
print(best_C)
t3 = time.time()
print("training time:", t3 - t2)
test_x, test_y = get_view_blocks(dist_matrices, test_inds, train_val_inds,
7), labels[test_inds]
pred = predict_svm_per_view(test_x, test_y, 7, models)
t4 = time.time()
print("testing time:", t4 - t3)
accuracies.append(accuracy_score(pred, test_y) * 100)
train_times.append(t3 - t2)
test_times.append(t4 - t3)
# acc_list.append(mean(mean_accuracies))
# std_list.append(stdev(mean_accuracies))
# train_time_list.append(mean(train_times))
# test_time_list.append(mean(test_times))
dict_to_csv(
{
'accuracy': mean(accuracies),
'error': stdev(accuracies),
'train_time': mean(train_times),
'test_time': mean(test_times)
}, ["nb_iter={},cv={}".format(1, 3)], PATH + ".csv")
t2 = time.time()
print("tuning time:", t2-t1)
# training
train_val_inds = np.hstack((train_inds,val_inds))
k_train_val_x = get_view_dict(k_x[np.ix_(train_val_inds,train_val_inds)])
model = train(k_train_val_x, y[train_val_inds], best_C)
t3 = time.time()
print("training time:", t3-t2)
test_inds = np.arange(len(test_y))+len(y)
k_test_x = get_view_dict(k_x[np.ix_(test_inds,train_val_inds)])
pred = predict(k_test_x, test_y, model)
t4 = time.time()
print("testing time:", t4-t3)
acc = accuracy_score(pred, test_y)*100
print(acc)
accuracies.append(acc)
times.append(t10-t0)
acc_list.append(mean(accuracies))
std_list.append(stdev(accuracies))
time_list.append(mean(times))
dict_to_csv({'accuracy':acc_list,'error':std_list,'times':time_list,'ratios':ratios_missing},["nb_iter={},cv={}".format(ITER,3)],PATH+".csv")
# training
train_val_x = get_view_dict(get_kernels(X, X, kernel=rbf_kernel))
mvml = one_vs_all_mvml_train(train_val_x, Y, 8, best_l, best_e, a)
t3 = time.time()
print("training time:", t3 - t2)
test_x = get_view_dict(get_kernels(test_X, X, kernel=rbf_kernel))
pred = one_vs_all_mvml_predict(test_x, mvml)
p_acc = accuracy_score(test_Y, pred)
t4 = time.time()
print("testing time:", t4 - t3)
acc_list.append(p_acc * 100)
std_list.append(0.)
train_time_list.append(t3 - t2)
test_time_list.append(t4 - t3)
dict_to_csv(
{
'accuracy': acc_list,
'error': std_list,
'train_time': train_time_list,
'test_time': test_time_list,
'rank': appr_levels
}, ["nb_iter={},cv={}".format(ITER, CV)], PATH + ".csv")