def convert_into_ellipses(data):
list_of_lists = []
for weights, atoms in data:
_list = []
for i, (center, sigma) in enumerate(atoms):
w, h, angle = get_ellipse_from_covariance(sigma)
e = patches.Ellipse(center, w, h, angle=angle, label='%d' % i)
e.set_alpha(weights[i])
_list.append(e)
list_of_lists.append(_list)
return list_of_lists
def original_clusters():
clusters = [[(1.5,2),((0.5,0.4),(0.4,0.5)),50000],
[(2,0),((0.3,0),(0,0.6)),30000],
[(4.5,1),((0.9,0.2),(0.2,0.3)),20000]]
artists = []
for center, cov, weight in clusters:
w, h, angle = get_ellipse_from_covariance(cov)
e = patches.Ellipse(center, w, h, angle=angle, color='r')
e.set_alpha(1. * weight / 60000)
artists.append(e)
return artists
def convert_into_ellipses(data):
list_of_lists = []
for weights, atoms in data:
_list = []
for i, (center, sigma) in enumerate(atoms):
w, h, angle = get_ellipse_from_covariance(sigma)
e = patches.Ellipse(center, w, h, angle=angle, label='%d' % i)
e.set_alpha(weights[i])
_list.append(e)
list_of_lists.append(_list)
return list_of_lists
def original_clusters():
clusters = [[(1.5, 2), ((0.5, 0.4), (0.4, 0.5)), 50000],
[(2, 0), ((0.3, 0), (0, 0.6)), 30000],
[(4.5, 1), ((0.9, 0.2), (0.2, 0.3)), 20000]]
artists = []
for center, cov, weight in clusters:
w, h, angle = get_ellipse_from_covariance(cov)
e = patches.Ellipse(center, w, h, angle=angle, color='r')
e.set_alpha(1. * weight / 60000)
artists.append(e)
return artists
# return '%s, %s\n%.2f %d' % (x[3]['controller_id'], x[3]['city'], x[2], x[1])
return '\tcid = {:10s}\n\tcity = {}\n\tratio = {:.2f}\n\tin cluster points = {:d}'.format(x[3]['controller_id'], x[3]['city'], x[2], x[1])
if __name__ == '__main__':
args = parse_args()
all_groups = extract_all_groups(args.log_file)
# with open('formateed_output.txt', 'w') as outfile:
# json.dump(all_groups, outfile)
centers = [tuple(x['data']) for x in json.load(open(args.centers))]
sigmas = load_cluster_parameters(args.covariances)
with open(args.weights) as infile:
weights = json.load(infile)
artists = []
orig_artists = []
for i in range(len(centers)):
w, h, angle = get_ellipse_from_covariance(sigmas[i])
e = patches.Ellipse(centers[i], w, h, angle=angle)
e_copy = patches.Ellipse(centers[i], w, h, angle=angle)
e.set_alpha(np.power(weights[i], .4))
e_copy.set_alpha(weights[i])
artists.append(e)
orig_artists.append(e_copy)
X, Y, Z = shape_data(args.scored_grid)
# fig = plt.figure()
# ax1 = fig.add_subplot(131)
# ax2 = fig.add_subplot(133)
fig, (ax1, ax2) = plt.subplots(1, 2, sharey=True)
fig.subplots_adjust(wspace=0)
plt.setp([a.get_yticklabels() for a in fig.axes[1:]], visible=False)
def plot_score_contours(args):
weights = []
centers = []
sigmas = []
if args.centers and args.covariances and args.weights:
# normalize weights to sum to 1.
weights = json.load(args.weights)
weights = [w / sum(weights) for w in weights]
weights = [0.4 * w / max(weights) for w in weights]
centers = load_cluster_parameters(args.centers)
sigmas = load_cluster_parameters(args.covariances)
fig = plt.figure(0)
ax = fig.add_subplot(111)
for i in range(len(centers)):
w, h, angle = get_ellipse_from_covariance(sigmas[i])
e = patches.Ellipse(centers[i], w, h, angle=angle)
e.set_alpha(weights[i])
ax.add_artist(e)
print i, weights[i], centers[i], sigmas[i]
set_ax_limits(ax, args)
x, y = zip(*centers)
plt.scatter(x, y, s=weights)
X, Y, Z = load_json_dump(args.scored_grid)
if args.score_cap:
Z = [min(z, args.score_cap) for z in Z]
if args.score_lower_limit:
Z = [max(z, args.score_lower_limit) for z in Z]
size = int(math.sqrt(len(Z)))
X = np.reshape(X, (size, size))
Y = np.reshape(Y, (size, size))
Z = np.reshape(Z, (size, size))
def format_args(i):
kwargs = {}
kwargs['mux'] = centers[i][0]
kwargs['muy'] = centers[i][1]
kwargs['sigmax'] = math.sqrt(sigmas[i][0][0])
kwargs['sigmay'] = math.sqrt(sigmas[i][1][1])
kwargs['sigmaxy'] = sigmas[i][0][1]
return kwargs
if len(weights):
Zgaussians = weights[0] * mlab.bivariate_normal(X, Y, **format_args(0))
for i in range(1, len(centers)):
Zgaussians += weights[i] * mlab.bivariate_normal(
X, Y, **format_args(i))
if args.plot == 'components':
CS = plt.contour(X, Y, Zgaussians, linewidth=10000, inline=1)
elif args.plot == 'density':
CS = plt.contour(X, Y, Z, linewidth=10000, inline=1)
elif args.plot == 'difference':
CS = plt.contour(X, Y, Z - Zgaussians, linewidth=10000, inline=1)
if args.plot != 'noop':
plt.clabel(CS, inline=1)
set_plot_limits(plt, args)
if args.csv and args.hist2d:
args.data = pd.read_csv(args.csv)
_plot_hist2d(args.data, args)
if args.savefig:
if args.savefig == SAVEFIG_INFER_VALUE:
name = os.path.basename(args.scored_grid).rsplit('.')[0]
filename = 'target/plots/{}.png'.format(name)
else:
filename = args.savefig
print 'saving figure to - ', filename
plt.savefig(filename, dpi=320)
else:
plt.show()
def plot_score_contours(args):
weights = []
centers = []
sigmas = []
if args.centers and args.covariances and args.weights:
# normalize weights to sum to 1.
weights = json.load(args.weights)
weights = [w / sum(weights) for w in weights]
weights = [0.4 * w / max(weights) for w in weights]
centers = load_cluster_parameters(args.centers)
sigmas = load_cluster_parameters(args.covariances)
fig = plt.figure(0)
ax = fig.add_subplot(111)
for i in range(len(centers)):
w, h, angle = get_ellipse_from_covariance(sigmas[i])
e = patches.Ellipse(centers[i], w, h, angle=angle)
e.set_alpha(weights[i])
ax.add_artist(e)
print i, weights[i], centers[i], sigmas[i]
set_ax_limits(ax, args)
x, y = zip(*centers)
plt.scatter(x, y, s=weights)
X, Y, Z = load_json_dump(args.scored_grid)
if args.score_cap:
Z = [min(z, args.score_cap) for z in Z]
if args.score_lower_limit:
Z = [max(z, args.score_lower_limit) for z in Z]
size = int(math.sqrt(len(Z)))
X = np.reshape(X, (size, size))
Y = np.reshape(Y, (size, size))
Z = np.reshape(Z, (size, size))
def format_args(i):
kwargs = {}
kwargs['mux'] = centers[i][0]
kwargs['muy'] = centers[i][1]
kwargs['sigmax'] = math.sqrt(sigmas[i][0][0])
kwargs['sigmay'] = math.sqrt(sigmas[i][1][1])
kwargs['sigmaxy'] = sigmas[i][0][1]
return kwargs
if len(weights):
Zgaussians = weights[0] * mlab.bivariate_normal(X, Y, **format_args(0))
for i in range(1, len(centers)):
Zgaussians += weights[i] * mlab.bivariate_normal(X, Y, **format_args(i))
if args.plot == 'components':
CS = plt.contour(X, Y, Zgaussians, linewidth=10000, inline=1)
elif args.plot == 'density':
CS = plt.contour(X, Y, Z, linewidth=10000, inline=1)
elif args.plot == 'difference':
CS = plt.contour(X, Y, Z - Zgaussians, linewidth=10000, inline=1)
if args.plot != 'noop':
plt.clabel(CS, inline=1)
set_plot_limits(plt, args)
if args.csv and args.hist2d:
args.data = pd.read_csv(args.csv)
_plot_hist2d(args.data, args)
if args.savefig:
if args.savefig == SAVEFIG_INFER_VALUE:
name = os.path.basename(args.scored_grid).rsplit('.')[0]
filename = 'target/plots/{}.png'.format(name)
else:
filename = args.savefig
print 'saving figure to - ', filename
plt.savefig(filename, dpi=320)
else:
plt.show()