scripts.plot.py Maven / Gradle / Ivy
import math
from matplotlib.pyplot import hist, title, subplot, scatter, plot
import matplotlib.pyplot as plt
import numpy as np
from PIL import Image
import seaborn # improves matplotlib look and feel
import sys
import time
'''
Optimization Methods Visualalization
Graph tools to help visualize how optimization is performing
'''
GLOBAL_TIME = 1.5
def load_file(path):
return np.loadtxt(path, delimiter=',')
def sigmoid(hidden_mean):
return 1 / (1 + np.exp(-hidden_mean))
def render_plot(values, plot_type='histogram', chart_title=''):
if np.product(values.shape) < 2:
values = np.zeros((3, 3))
chart_title += '-fake'
if plot_type == 'histogram':
hist(values)
elif plot_type == "scatter":
scatter(values)
else:
print "The " + plot_type + " format is not supported. Please choose histogram or scatter."
magnitude = ' mm %g ' % np.mean(np.fabs(values))
chart_title += ' ' + magnitude
title(chart_title)
def render_activation_probability(dataPath, filename):
hidden_mean = load_file(dataPath)
img = Image.fromarray(sigmoid(hidden_mean) * 256)
if img.mode != 'RGB':
img = img.convert('RGB')
img.save(filename, 'PNG')
def plot_single_graph(path, chart_title, filename):
print 'Graphing ' + chart_title + '\n'
values = load_file(path)
plt.plot(values, 'b')
plt.title(chart_title)
plt.savefig(filename, format='png')
plt.show(block=False)
time.sleep(GLOBAL_TIME)
plt.close()
def plot_matrices(orig_path, plot_type, filename):
paths = orig_path.split(',')
for idx, path in enumerate(paths):
if idx % 2 == 0:
title = paths[idx + 1]
print 'Loading matrix ' + title + '\n'
matrix = load_file(path)
subplot(2, len(paths)/4, idx/2+1)
render_plot(matrix, plot_type, chart_title=title)
plt.tight_layout()
plt.savefig(filename, format='png')
plt.show(block=False)
time.sleep(GLOBAL_TIME)
plt.close()
# TODO Finish adapting. Code still does not fully run through.
# def render_filter(data_path, n_rows, n_cols, filename):
# weight_data = load_file(data_path).reshape((n_rows, n_cols))
# patch_width = weight_data.shape[1]
# patch_height = 1
#
# # Initialize background to dark gray
# filter_frame = np.ones((n_rows*patch_width, n_cols * patch_height), dtype='uint8')
#
# for row in xrange(int(n_rows/n_cols)):
# for col in xrange(n_cols):
# patch = weight_data[row * n_cols + col].reshape((patch_width, patch_height))
# norm_patch = ((patch - patch.min()) / (patch.max() - patch.min() + 1e-6))
# filter_frame[row * patch_width: row * patch_width + patch_width,
# col * patch_height:col * patch_height + patch_height] = norm_patch * 255
# img = Image.fromarray(filter_frame)
# img.savefig(filename)
# img.show()
#
# def render_filter(data_path, filename, filter_width=10, filter_height=10):
# print 'Rendering filter image...'
# weight_data = load_file(data_path)
# n_rows = weight_data.shape[0]
# n_cols = weight_data.shape[1]
# padding = 1
#
# # Initialize background to dark gray
# filter_frame = np.ones(((filter_width+padding) * filter_width, (filter_height+padding) * filter_height), dtype='uint8') * 51
#
# for row in xrange(n_rows):
# for col in xrange(n_cols):
# patch = weight_data[row * n_cols + col].reshape((filter_width, filter_height))
# norm_patch = ((patch - patch.min()) / (patch.max() - patch.min() + 1e-6))
# filter_frame[row * (filter_height+padding): row * (filter_height+padding)+filter_height, col * (filter_width+padding): col * (filter_width+padding)+filter_width] = norm_patch * 255
# filter_frame[row * (filter_height+padding): row * (filter_height+padding) + filter_height, col * (filter_width+padding): col *(filter_width+padding) + filter_width]
# img = Image.fromarray(filter_frame)
# if img.mode != 'RGB':
# img = img.convert('RGB')
# img.save(filename)
# def vis_square(data_path, filename, n_rows=28, n_cols=28, padsize=1, padval=0):
# data = load_file(data_path)
# data = data.reshape(n_rows, n_cols)
#
# data -= data.min()
# data /= data.max()
#
# # force the number of filters to be square
# n = int(np.ceil(np.sqrt(data.shape[0])))
# padding = ((0, n ** 2 - data.shape[0]), (0, padsize), (0, padsize)) + ((0, 0),) * (data.ndim - 3)
# data = np.pad(data, padding, mode='constant', constant_values=(padval, padval))
#
# # tile the filters into an image
# data = data.reshape((n, n) + data.shape[1:]).transpose((0, 2, 1, 3) + tuple(range(4, data.ndim + 1)))
# data = data.reshape((n * data.shape[1], n * data.shape[3]) + data.shape[4:])
#
# plt.imshow(data)
# time.sleep(GLOBAL_TIME)
# plt.savefig(data, filename)
if __name__ == '__main__':
if len(sys.argv) < 4:
print 'Please specify a command: One of hbias,weights,plot and a file path'
sys.exit(1)
plot_type = sys.argv[1]
path = sys.argv[2]
filename = sys.argv[3]
if plot_type == 'activations':
render_activation_probability(path, filename)
elif plot_type == 'single_matrix':
render_plot(path)
elif plot_type == 'histogram':
plot_matrices(path, plot_type, filename)
elif plot_type == 'scatter':
plot_matrices(path, plot_type, filename)
elif plot_type == 'loss':
plot_single_graph(path, plot_type, filename)
elif plot_type == 'accuracy':
plot_single_graph(path, plot_type, filename)
# elif sys.argv[1] == 'filter':
# if sys.argv[7]:
# n_rows = int(sys.argv[4])
# n_cols = int(sys.argv[5])
# filter_width = int(sys.argv[6])
# filter_height = int(sys.argv[7])
# render_filter(path, filename, n_rows, n_cols, filter_height, filter_width)
# elif sys.argv[5]:
# n_rows = int(sys.argv[4])
# n_cols = int(sys.argv[5])
# render_filter(path, filename, n_rows, n_cols)
# else:
# render_filter(path, filename)
© 2015 - 2025 Weber Informatics LLC | Privacy Policy