"""
import matplotlib.pyplot as plot
plot.figure(random.randint(0, 10000000))
plot.scatter(data[0], data[1], 20, 'b', 'o')
plot.title(filename.split('.')[0])
for line in lines:
plot.plot([line[0], line[2]], [line[1], line[3]], '-')
plot.savefig(filename)
def T(a):
return a.reshape(len(a), 1)
x = load('pcaData.txt')
U, s, x_rot = pca.pca(x)
scatter('raw-scatterplot.png', x, [((0, 0) + tuple(r)) for r in U.T])
scatter('x-rot.png', x_rot)
k = 1
U, s, xHat = pca.pca(x, k)
scatter('xHat.png', np.vstack([xHat, xHat]))
xPCAwhite = pca.pca_whiten(x)
scatter('xPCAwhite.png', xPCAwhite)
xZCAwhite = pca.zca_whiten(x)
scatter('xZCAwhite.png', xZCAwhite)
x_hat = np.dot(U[:, :k], small_x_rot)
display_network.display_network('99-reduced.png', x_hat)
# first index greater than 90%
k = np.min(np.where(pov >= 0.90))
print 'k:', k
U, s, small_x_rot = pca.pca(patches, k=k)
x_hat = np.dot(U[:, :k], small_x_rot)
display_network.display_network('90-reduced.png', x_hat)
# first index greater than 50%
k = np.min(np.where(pov >= 0.50))
print 'k:', k
U, s, small_x_rot = pca.pca(patches, k=k)
x_hat = np.dot(U[:, :k], small_x_rot)
display_network.display_network('50-reduced.png', x_hat)
epsilon = 0.1
x_pca_white = pca.pca_whiten(patches, epsilon=epsilon)
c = pca.covariance(x_pca_white)
display_network.array_to_file('pca_white_covariance.png', c)
epsilon = 0.0
x_pca_white = pca.pca_whiten(patches, epsilon=epsilon)
c = pca.covariance(x_pca_white)
display_network.array_to_file('pca_white_noreg_covariance.png', c)
for epsilon in [1, 0.1, 0.01]:
x_zca_white = pca.zca_whiten(patches, epsilon=epsilon)
display_network.display_network('zca_%s.png' % epsilon, x_zca_white)
k = np.min(np.where(pov >= 0.90))
print 'k:', k
U, s, small_x_rot = pca.pca(patches, k=k)
x_hat = np.dot(U[:,:k], small_x_rot)
display_network.display_network('90-reduced.png', x_hat)
# first index greater than 50%
k = np.min(np.where(pov >= 0.50))
print 'k:', k
U, s, small_x_rot = pca.pca(patches, k=k)
x_hat = np.dot(U[:,:k], small_x_rot)
display_network.display_network('50-reduced.png', x_hat)
epsilon = 0.1
x_pca_white = pca.pca_whiten(patches, epsilon=epsilon)
c = pca.covariance(x_pca_white)
display_network.array_to_file('pca_white_covariance.png', c)
epsilon = 0.0
x_pca_white = pca.pca_whiten(patches, epsilon=epsilon)
c = pca.covariance(x_pca_white)
display_network.array_to_file('pca_white_noreg_covariance.png', c)
for epsilon in [1, 0.1, 0.01]:
x_zca_white = pca.zca_whiten(patches, epsilon=epsilon)
display_network.display_network('zca_%s.png' % epsilon, x_zca_white)
"""
import matplotlib.pyplot as plot
plot.figure(random.randint(0, 10000000))
plot.scatter(data[0], data[1], 20, 'b', 'o')
plot.title(filename.split('.')[0])
for line in lines:
plot.plot([line[0], line[2]], [line[1], line[3]], '-')
plot.savefig(filename)
def T(a):
return a.reshape(len(a), 1)
x = load('pcaData.txt')
U, s, x_rot = pca.pca(x)
scatter('raw-scatterplot.png', x, [((0,0)+tuple(r)) for r in U.T])
scatter('x-rot.png', x_rot)
k = 1
U, s, xHat = pca.pca(x, k)
scatter('xHat.png', np.vstack([xHat, xHat]))
xPCAwhite = pca.pca_whiten(x)
scatter('xPCAwhite.png', xPCAwhite)
xZCAwhite = pca.zca_whiten(x)
scatter('xZCAwhite.png', xZCAwhite)