def wheel_pass_debugging(road: Road, wheel: Wheel, max_iterations: int,
bump_method: str, dig_method: str,
dig_probability_arguments: list,
smoothing_method: str, smoothing_arguments: list):
"""
This function performs a loop of wheel passes through a road until 'max_iterations'
wheel passes have been performed. The successive wheel passes modify the road surface.
:param road: a Road
:param wheel: a Wheel
:param max_iterations: maximum number of wheel passes (iterations)
:param bump_method: str, method name for the 'determine_bump_height' function
:param dig_method: str, method name for the 'digging' function
:param dig_probability_arguments:
:param smoothing_method:
:param smoothing_arguments:
"""
current_passes = 0
while wheel.number_of_passes < max_iterations:
if current_passes < wheel.number_of_passes:
smoothing(road, wheel, smoothing_method, smoothing_arguments)
current_passes = wheel.number_of_passes
initial_position = wheel.xf
print_road_surface(road, wheel.xf, wheel.diameter)
bump_position = wheel.move_to_next_bump(road)
bump_height = determine_bump_height(road,
wheel,
bump_position,
method=bump_method)
print_road_surface(road, wheel.xf, wheel.diameter)
wheel.jump(road, bump_height)
print_road_surface(road, wheel.xf, wheel.diameter)
digging(road,
wheel,
wheel.xf,
method=dig_method,
dig_probability_args=dig_probability_arguments)
wheel.update_position(wheel.diameter)
wheel.set_elevation(road.piles[wheel.xf])
final_position = wheel.xf
if final_position <= initial_position:
smoothing(road, wheel, smoothing_method, smoothing_arguments)
wheel.number_of_passes += 1
print(f'\nIteration number {wheel.number_of_passes}')
print(
f'The number of grains is {road.get_number_of_grains()}, the initial was {road.initial_number_of_grains}\n'
)
print_road_surface(road, wheel.xf, wheel.diameter)
def online_train(step=1):
start_time = time.time()
data_labels, tag_map = create_matrix.get_data_labels_and_tag_map()
fm_obj = fm_online.FmOnline(data_labels, tag_map)
fm_obj.prepare_train(0.005, K=8, step=1)
fm_obj.fit(step)
fm_obj.calc_error()
fm_obj.arrange_user()
redis_flush()
fm_obj.cy_fm.save_redis()
labels = fm_obj.labels
save_params_into_radis(labels, tag_map) # labelsをredisに保存
smoothing()
fm_obj.save_top_k_ranking_all_user(smoothing_flag = True)
print time.time() - start_time
def smoothing_validation():
start_time = time.time()
learning_matrix, regs_matrix, labels, targets, tag_map, ratelist, train_songs, validation_songs = create_matrix.create_smoothing_fm_matrix()
redis_flush(db=1)
save_songs("train_songs", train_songs)
save_songs("validation_songs", validation_songs)
print "FMクラス初期化"
fm_obj = fm_batch.FmBatch(learning_matrix, regs_matrix, labels, targets, tag_map)
print "SGDで学習開始"
fm_obj.learning(0.005, K=8, step=1)
fm_obj.arrange_user()
fm_obj.cy_fm.save_redis(db=1)
labels = fm_obj.labels
save_params_into_radis(labels, tag_map) # labelsをredisに保存
#fm_obj.smoothing(smoothing_evaluate=True)
smoothing(smoothing_evaluate=True)
print time.time() - start_time
def batch_train():
start_time = time.time()
learning_matrix, regs_matrix, labels, targets, tag_map, ratelist = create_matrix.create_fm_matrix()
print "FMクラス初期化"
fm_obj = fm_batch.FmBatch(learning_matrix, regs_matrix, labels, targets, tag_map)
print "SGDで学習開始"
fm_obj.learning(0.005, K=8, step=1)
fm_obj.arrange_user()
#fm_obj.smoothing()
print "redisに保存"
redis_flush()
fm_obj.cy_fm.save_redis()
labels = fm_obj.labels
save_params_into_radis(labels, tag_map) # labelsをredisに保存
#print "top_k_ranking保存"
#fm_obj.save_top_k_ranking_all_user()
smoothing()
print "top_k_ranking保存"
fm_obj.save_top_k_ranking_all_user(smoothing_flag = True)
print time.time() - start_time
# Image shape (91, 109, 91, 240)
#in_brain_img = make_mask_filtered_data(image_path[1],mask_path)
data_int = in_brain_img.get_data()
data = data_int.astype(float)
mean_data = np.mean(data, axis=-1)
in_brain_mask = (mean_data - 0.0) < 0.01
Transpose = False
else:
img = nib.load(image_path[1])
data_int = img.get_data()
data = data_int.astype(float)
mean_data = np.mean(data, axis=-1)
in_brain_mask = mean_data > thres
Transpose = True
# Smoothing with Gaussian filter
smooth_data = smoothing(data,1,range(data.shape[-1]))
# Selecting the voxels in the brain
in_brain_tcs = smooth_data[in_brain_mask, :]
#in_brain_tcs = data[in_brain_mask, :]
vol_shape = data.shape[:-1]
# Plotting the voxels in the brain
plt.imshow(plot_mosaic(mean_data, transpose=Transpose), cmap='gray', alpha=1)
plt.colorbar()
plt.contour(plot_mosaic(in_brain_mask, transpose=Transpose),colors='blue')
plt.title('In brain voxel mean values' + '\n' + (d_path['type'] + str(name)))
plt.savefig(project_path+'fig/BOLD/%s_mean_voxels_countour.png'\
%(d_path['type'] + str(name)))
#plt.show()
plt.clf()
template_data = template_data_int.astype(float)
Transpose = False
in_brain_mask = (mean_data - 0.0) < 0.01
plt.imshow(plot_mosaic(template_data, transpose=Transpose),\
cmap='gray', alpha=1)
else:
img = nib.load(image_path[1])
data_int = img.get_data()
data = data_int.astype(float)
mean_data = np.mean(data, axis=-1)
in_brain_mask = mean_data > thres
Transpose = True
plt.contour(plot_mosaic(in_brain_mask, transpose=Transpose), \
cmap='gray' , alpha=1)
# Smoothing with Gaussian filter
smooth_data = smoothing(data,1,range(data.shape[-1]))
# Selecting the voxels in the brain
in_brain_tcs = smooth_data[in_brain_mask, :]
#in_brain_tcs = data[in_brain_mask, :]
vol_shape = data.shape[:-1]
# Plotting the voxels in the brain
plt.imshow(plot_mosaic(mean_data, transpose=Transpose), cmap='gray', alpha=1)
plt.colorbar()
plt.title('In brain voxel mean values' + '\n' + (d_path['type'] + str(name)))
plt.savefig(project_path+'fig/BOLD/%s_mean_voxels.png'\
%(d_path['type'] + str(name)))
#plt.show()
#plt.clf()
# Convolution with 1 to 4 conditions
X_matrix_high_res1 = np.loadtxt(txt_path[6])
X_matrix_high_res2 = np.loadtxt(txt_path[7])
X_matrix_high_res3 = np.loadtxt(txt_path[8])
X_matrix_high_res4 = np.loadtxt(txt_path[9])
X_matrix_high_res = np.ones((len(X_matrix1), p))
X_matrix_high_res[..., 1] = X_matrix_high_res1
X_matrix_high_res[..., 2] = X_matrix_high_res2
X_matrix_high_res[..., 3] = X_matrix_high_res3
X_matrix_high_res[..., 4] = X_matrix_high_res4
beta_4d = glm_beta(data, X_matrix)
MRSS, fitted, residuals = glm_mrss(beta_4d, X_matrix, data)
# smooth the data and re-run the regression
data_smooth = smoothing(data, 1, range(data.shape[-1]))
beta_4d_smooth = glm_beta(data_smooth, X_matrix)
MRSS_smooth, fitted_smooth, residuals_smooth = glm_mrss(
beta_4d_smooth, X_matrix, data_smooth)
# use high resolution to create our design matrix
beta_4d_high_res = glm_beta(data, X_matrix_high_res)
MRSS_high_res, fitted_high_res, residuals_high_res = glm_mrss(
beta_4d_high_res, X_matrix_high_res, data)
plt.plot(data[4, 22, 11], label="actual")
plt.plot(fitted[4, 22, 11], label="fitted")
plt.plot(fitted_high_res[4, 22, 11], label="fitted_high_res")
plt.title(name[0:17] + "voxel (4,22,11) actual vs fitted")
plt.legend(loc="upper left", fontsize="smaller")
X_matrix_high_res1 = np.loadtxt(txt_path[6])
X_matrix_high_res2 = np.loadtxt(txt_path[7])
X_matrix_high_res3 = np.loadtxt(txt_path[8])
X_matrix_high_res4 = np.loadtxt(txt_path[9])
X_matrix_high_res = np.ones((len(X_matrix1),p))
X_matrix_high_res[...,1] = X_matrix_high_res1
X_matrix_high_res[...,2] = X_matrix_high_res2
X_matrix_high_res[...,3] = X_matrix_high_res3
X_matrix_high_res[...,4] = X_matrix_high_res4
beta_4d = glm_beta(data,X_matrix)
MRSS, fitted, residuals = glm_mrss(beta_4d, X_matrix, data)
# smooth the data and re-run the regression
data_smooth = smoothing(data,1,range(data.shape[-1]))
beta_4d_smooth = glm_beta(data_smooth,X_matrix)
MRSS_smooth, fitted_smooth, residuals_smooth = glm_mrss(beta_4d_smooth, X_matrix, data_smooth)
# use high resolution to create our design matrix
beta_4d_high_res = glm_beta(data,X_matrix_high_res)
MRSS_high_res, fitted_high_res, residuals_high_res = glm_mrss(beta_4d_high_res, X_matrix_high_res, data)
plt.plot(data[4,22,11], label = "actual")
plt.plot(fitted[4,22,11], label = "fitted")
plt.plot(fitted_high_res[4,22,11], label = "fitted_high_res")
plt.title(name[0:17]+"voxel (4,22,11) actual vs fitted")
plt.legend(loc = "upper left", fontsize = "smaller")
plt.savefig(dirs[1] + '/'+ name[0:17]+ "_glm_fitted.png")
thres = 375 #from analysis of the histograms
for image_path in images_paths:
name = image_path[0]
print("Starting t-test analysis and plot for subject "+name[9:12])
img = nib.load(image_path[1])
data_int = img.get_data()
data = data_int.astype(float)
vol_shape = data.shape[:-1]
n_trs = data.shape[-1]
#get the mean value
mean_data = np.mean(data, axis = -1)
#build the mask
in_brain_mask = mean_data > 375
#smooth the data set
smooth_data = smoothing(data, 1, range(n_trs))
#initialize design matrix for t test
p = 7
X_matrix = np.ones((data.shape[-1], p))
#build our design matrix
for cond in range(1,5):
convolved = np.loadtxt(txt_path + name + '_conv_' + str(cond).zfill(3) + '_high_res.txt')
#convolved = np.loadtxt(txt_path + name + '_conv_' + str(cond).zfill(3) + '_canonical.txt')
X_matrix[:,cond] = convolved
linear_drift = np.linspace(-1, 1, n_trs)
X_matrix[:,5] = linear_drift
quadratic_drift = linear_drift ** 2
quadratic_drift -= np.mean(quadratic_drift)
X_matrix[:,6] = quadratic_drift
beta, t, df, p = t_stat(smooth_data, X_matrix)
for cond in range(0,4):
