def plot1(gp, ngrid=100, lim=None, k=range(-3, 4)):
k = sorted(k)
if lim is None:
lim = (np.amin(gp.x[:, 1]), np.amax(gp.x[:, 1]))
x = np.linspace(lim[0], lim[1], ngrid).T
(m, v) = gp.inf(x)
m = np.asarray(m).squeeze()
v = np.asarray(v).squeeze()
plt.plot(x, m,
color=DARKBLUE,
linewidth=2)
for i in k:
if i == 0: continue
lo = m - i*np.sqrt(v)
hi = m + i*np.sqrt(v)
plt.fill_between(x, lo, hi,
linestyle='solid',
edgecolor=DARKGRAY,
facecolor=LIGHTGRAY,
alpha=0.2)
plt.plot(gp.x, gp.y,
'o',
markersize=8,
markeredgewidth=1,
markeredgecolor=DARKGRAY,
markerfacecolor=LIGHTBLUE)
plt.xlim(lim)
def __init__(self, trainingSet, outputFile):
self.trainingSet_ = trainingSet
self.outputFile_ = outputFile
# initialize weights
self.weights1_ = matlib.rand(self.HIDDEN, self.FEATURES)
self.weights1_ = self.weights1_ / matlib.sqrt(self.FEATURES)
self.weights2_ = matlib.rand(self.FEATURES, self.HIDDEN)
self.weights2_ = self.weights2_ / matlib.sqrt(self.HIDDEN)
# initialize bias
self.bias1_ = matlib.zeros((self.HIDDEN, ))
self.bias2_ = matlib.zeros((self.FEATURES, ))
# initialize rho estimate vector
self.rho_est_ = matlib.zeros((self.HIDDEN, )).T
def auclidic_distance(x: pn.DataFrame, y: pn.DataFrame) -> float:
sum = 0
params = [
'Pregnancies', 'Glucose', 'BloodPressure', 'SkinThickness',
'Insulin', 'BMI', 'DiabetesPedigreeFunction', 'Age'
]
for p in params:
sum += abs(x[p] - y[p])**2
return np.sqrt(sum)
def __init_learner(self, outputFile):
self.outputFile_ = outputFile
# initialize weights
self.weights1_ = matlib.rand(self.HIDDEN, self.FEATURES)
self.weights1_ = self.weights1_ / matlib.sqrt(self.FEATURES)
self.weights2_ = matlib.rand(self.FEATURES, self.HIDDEN)
self.weights2_ = self.weights2_ / matlib.sqrt(self.HIDDEN)
# initialize bias
self.bias1_ = matlib.zeros((self.HIDDEN, ))
self.bias2_ = matlib.zeros((self.FEATURES, ))
# initialize rho estimate vector
self.rho_est_ = matlib.zeros((self.HIDDEN, )).T
self.errors = []
def __init_learner(self, outputFile):
self.outputFile_ = outputFile
# initialize weights
self.weights1_ = matlib.rand(self.HIDDEN, self.FEATURES)
self.weights1_ = self.weights1_ / matlib.sqrt(self.FEATURES)
self.weights2_ = matlib.rand(self.FEATURES, self.HIDDEN)
self.weights2_ = self.weights2_ / matlib.sqrt(self.HIDDEN)
# initialize bias
self.bias1_ = matlib.zeros((self.HIDDEN,))
self.bias2_ = matlib.zeros((self.FEATURES,))
# initialize rho estimate vector
self.rho_est_ = matlib.zeros((self.HIDDEN,)).T
self.errors = []
def __init_projector(self, inputFile):
handle = open(inputFile, "r+")
# initialize weights
self.weights1_ = matlib.zeros((self.HIDDEN, self.FEATURES))
self.weights1_ = self.weights1_ / matlib.sqrt(self.FEATURES)
for r, line in enumerate(handle.readlines()):
for c, col in enumerate(line.strip().split(" ")):
self.weights1_[r, c] = float(col)
handle.close()
# initialize bias
self.bias1_ = matlib.zeros((self.HIDDEN, ))
def plot_cov(mu, sigma):
mu = np.asarray(mu)
sigma = np.asarray(sigma)
t = np.linspace(-math.pi, math.pi, 2*math.pi/.1)
x = np.sin(t)
y = np.cos(t)
D_diag, V = ml.linalg.eigh(sigma)
D = np.diag(D_diag)
A = np.real((np.dot(V,ml.sqrt(D))).T)
z = np.dot(np.vstack((x.T,y.T)).T,A)
plt.plot(z[:,0]+mu[0], z[:,1]+mu[1], 'y-')
def plot_cov(mu, sigma, plotType = 'b-', alpha=1):
mu = np.asarray(mu)
sigma = np.asarray(sigma)
t = np.linspace(-np.pi, np.pi, 2*np.pi/.1)
x = np.sin(t)
y = np.cos(t)
D_diag, V = ml.linalg.eigh(sigma)
D = np.diag(D_diag)
A = np.real((np.dot(V,ml.sqrt(D))).T)
z = np.dot(np.vstack((x.T,y.T)).T,A)
plt.plot(z[:,0]+mu[0], z[:,1]+mu[1], plotType, alpha=alpha, linewidth=4, mew=2.0)
def plot_cov(mu, sigma, plotType='b-', alpha=1):
mu = np.asarray(mu)
sigma = np.asarray(sigma)
t = np.linspace(-math.pi, math.pi, 2 * math.pi / .1)
x = np.sin(t)
y = np.cos(t)
D_diag, V = ml.linalg.eigh(sigma)
D = np.diag(D_diag)
A = np.real((np.dot(V, ml.sqrt(D))).T)
z = np.dot(np.vstack((x.T, y.T)).T, A)
plt.plot(z[:, 0] + mu[0], z[:, 1] + mu[1], plotType, alpha=alpha)
def __init_projector(self, inputFile):
handle = open(inputFile, "r+")
# initialize weights
self.weights1_ = matlib.zeros((self.HIDDEN, self.FEATURES))
self.weights1_ = self.weights1_ / matlib.sqrt(self.FEATURES)
for r, line in enumerate(handle.readlines()):
for c, col in enumerate(line.strip().split(" ")):
self.weights1_[r, c] = float(col)
handle.close()
# initialize bias
self.bias1_ = matlib.zeros((self.HIDDEN,))
def auclidic_distance(x: pn.DataFrame, y: pn.DataFrame,
subset: list) -> float:
sum = 0
params = [
'Pregnancies', 'Glucose', 'BloodPressure', 'SkinThickness',
'Insulin', 'BMI', 'DiabetesPedigreeFunction', 'Age'
]
modified_params = []
if subset is None:
modified_params = params
elif len(subset) == 1:
modified_params.append(params[subset[0]])
else:
for idx in subset:
modified_params.append(params[idx])
for p in modified_params:
sum += abs(x[p] - y[p])**2
return np.sqrt(sum)
def return_magnitude(self):
"""Returns the magnitude matrix of this matrixvector"""
from numpy.matlib import sqrt
return sqrt(elementwise_dot_product(self, self))
def rand_init(fan_in, fan_out):
ret = np.asarray(rng.uniform(low=-np.sqrt(3. / fan_in),
high=np.sqrt(3. / fan_in),
size=(fan_in, fan_out)),
dtype=T.config.floatX)
return ret
def haar_matrix(dim):
'''
Generate haar transformation matrix.
'''
Q = M.matrix('1 1;-1 1')
return 1 / M.sqrt(2) * M.kron(M.eye(dim / 2), Q)
def scaleParams(self):
for con in self.network._containerIterator():
factor = 1.0 / sqrt(con.indim)
for param in range(len(con.params)):
con.params[param] *= factor
def rand_init(fan_in, fan_out):
ret = np.asarray(rng.uniform(
low = -np.sqrt(3. / fan_in),
high = np.sqrt(3. / fan_in),
size = (fan_in, fan_out)), dtype = T.config.floatX)
return ret
def haar_matrix(dim):
'''
Generate haar transformation matrix.
'''
Q = M.matrix('1 1;-1 1')
return 1 / M.sqrt(2) * M.kron(M.eye(dim / 2), Q)
def return_magnitude(self) -> matrix:
"""Returns the magnitude matrix of this matrix vector"""
return sqrt(square(self.x) + square(self.y))
