def parcel_plv(x, y, source_identities):
""" Function for computing the complex phase-locking value at parcel level.
Input:
x : complex Source time series.
y : complex Source time series.
source_identities : int Vector mapping of source parcel identities.
Output:
cPLV: complex PLV of parcels, sorted by identity.
"""
"""Change to amplitude 1, keep angle using Euler's formula."""
x = np.exp(1j * (asmatrix(np.angle(x))))
y = np.exp(1j * (asmatrix(np.angle(y))))
"""Get cPLV needed for flips and weighting."""
cplv = np.zeros(len(source_identities), dtype='complex')
for i, identity in enumerate(source_identities):
"""Compute cPLV only of parcel source pairs of sources that
belong to that parcel. One source belong to only one parcel."""
if (source_identities[i] >= 0):
cplv[i] = (np.sum(
(np.asarray(y[identity])) * np.conjugate(np.asarray(x[i]))))
cplv /= np.shape(x)[1]
return cplv
def load_data(fname, delimiter=','):
""" return the features x and result y as matrix """
data = sp.loadtxt(fname, delimiter=delimiter)
m, n = data.shape
x = sp.asmatrix(data[:, range(0, n - 1)].reshape(m, n - 1))
y = sp.asmatrix(data[:, n - 1].reshape(m, 1))
return x, y
def dare(A, B, Q, R, S=None, E=None, stabilizing=True):
""" (X,L,G) = dare(A,B,Q,R) solves the discrete-time algebraic Riccati
equation
:math:`A^T X A - X - A^T X B (B^T X B + R)^{-1} B^T X A + Q = 0`
where A and Q are square matrices of the same dimension. Further, Q
is a symmetric matrix. The function returns the solution X, the gain
matrix G = (B^T X B + R)^-1 B^T X A and the closed loop eigenvalues L,
i.e., the eigenvalues of A - B G.
(X,L,G) = dare(A,B,Q,R,S,E) solves the generalized discrete-time algebraic
Riccati equation
:math:`A^T X A - E^T X E - (A^T X B + S) (B^T X B + R)^{-1} (B^T X A + S^T) + Q = 0`
where A, Q and E are square matrices of the same dimension. Further, Q and
R are symmetric matrices. The function returns the solution X, the gain
matrix :math:`G = (B^T X B + R)^{-1} (B^T X A + S^T)` and the closed loop
eigenvalues L, i.e., the eigenvalues of A - B G , E.
"""
if S is not None or E is not None or not stabilizing:
return dare_old(A, B, Q, R, S, E, stabilizing)
else:
Rmat = asmatrix(R)
Qmat = asmatrix(Q)
X = solve_discrete_are(A, B, Qmat, Rmat)
G = solve(B.T.dot(X).dot(B) + Rmat, B.T.dot(X).dot(A))
L = eigvals(A - B.dot(G))
return X, L, G
def dare(A, B, Q, R, S=None, E=None):
""" (X,L,G) = dare(A,B,Q,R) solves the discrete-time algebraic Riccati
equation
A^T X A - X - A^T X B (B^T X B + R)^-1 B^T X A + Q = 0
where A and Q are square matrices of the same dimension. Further, Q
is a symmetric matrix. The function returns the solution X, the gain
matrix G = (B^T X B + R)^-1 B^T X A and the closed loop eigenvalues L,
i.e., the eigenvalues of A - B G.
(X,L,G) = dare(A,B,Q,R,S,E) solves the generalized discrete-time algebraic
Riccati equation
A^T X A - E^T X E - (A^T X B + S) (B^T X B + R)^-1 (B^T X A + S^T) +
+ Q = 0
where A, Q and E are square matrices of the same dimension. Further, Q and
R are symmetric matrices. The function returns the solution X, the gain
matrix G = (B^T X B + R)^-1 (B^T X A + S^T) and the closed loop
eigenvalues L, i.e., the eigenvalues of A - B G , E.
"""
if S is not None or E is not None:
return dare_old(A, B, Q, R, S, E)
else:
Rmat = asmatrix(R)
Qmat = asmatrix(Q)
X = solve_discrete_are(A, B, Qmat, Rmat)
G = inv(B.T.dot(X).dot(B) + Rmat) * B.T.dot(X).dot(A)
L = eigvals(A - B.dot(G))
return X, L, G
def rows_array2(my_m, n_iter):
my_m = asmatrix(my_m) # not strictly needed, but should be noop
# the following is valid as long as we assume a fixed matrix size (and not in a general function)
v1 = asmatrix(full((my_m.shape[1], 1), 1.0 / my_m.shape[1]))
for i in xrange(n_iter):
#v1 = asmatrix(full( (my_m.shape[1], 1), 1.0/my_m.shape[1]))
m1 = my_m * v1 # row means
m1[0] = 0 # just to avoid complaints
def count_hops(data, definition, def_args, year, A, B):
scipy_matrix = scipy.asmatrix(scipy.array(unsigned_adjacency_matrix(data, definition, def_args, year)))
multiplied_matrix = scipy.asmatrix(scipy.array(unsigned_adjacency_matrix(data, definition, def_args, year)))
hop_count = 1
while hop_count < len(data.countries()):
if multiplied_matrix.tolist()[index_of_country(A)][index_of_country(B)] != 0:
return hop_count
multiplied_matrix = memoize_matrix_mult(multiplied_matrix, scipy_matrix)
hop_count += 1
return INFINITE_HOPS
def rankOneMatrix(vec1, *args):
"""
Create rank one matrices (dyadics) from vectors.
r1mat = rankOneMatrix(vec1)
r1mat = rankOneMatrix(vec1, vec2)
vec1 is m1 x n, an array of n hstacked m1 vectors
vec2 is m2 x n, (optional) another array of n hstacked m2 vectors
r1mat is n x m1 x m2, an array of n rank one matrices
formed as c1*c2' from columns c1 and c2
With one argument, the second vector is taken to
the same as the first.
Notes:
*) This routine loops on the dimension m, assuming this
is much smaller than the number of points, n.
"""
if len(vec1.shape) > 2:
raise RuntimeError("input vec1 is the wrong shape")
if (len(args) == 0):
vec2 = vec1.copy()
else:
vec2 = args[0]
if len(vec1.shape) > 2:
raise RuntimeError("input vec2 is the wrong shape")
m1, n1 = asmatrix(vec1).shape
m2, n2 = asmatrix(vec2).shape
if (n1 != n2):
raise RuntimeError("Number of vectors differ in arguments.")
m1m2 = m1 * m2
r1mat = zeros((m1m2, n1), dtype='float64')
mrange = asarray(list(range(m1)), dtype='int')
for i in range(m2):
r1mat[mrange, :] = vec1 * tile(vec2[i, :], (m1, 1))
mrange = mrange + m1
r1mat = reshape(r1mat.T, (n1, m2, m1)).transpose(0, 2, 1)
return squeeze(r1mat)
def main():
saved_handler = sp.seterrcall(err_handler)
saved_err = sp.seterr(all='call')
print('============ Part 1: Plotting =============================')
x, y = load_data('ex2/ex2data1.txt')
plot_data(x, y)
pl.show()
print('============ Part 2: Compute Cost and Gradient ============')
m, n = x.shape
x = sp.column_stack((sp.ones((m, 1)), x))
init_theta = sp.asmatrix(sp.zeros((n + 1, 1)))
cost, grad = cost_function(init_theta, x, y)
print('Cost at initial theta: %s' % cost)
print('Gradient at initial theta:\n %s' % grad)
print('============ Part 3: Optimizing minimize ====================')
# res = op.minimize(cost_function, init_theta, args=(x, y), jac=True, method='Newton-CG')
res = op.minimize(cost_function_without_grad, init_theta, args=(x, y), method='Powell')
# print('Cost at theta found by fmin: %s' % cost)
print('Result by minimize:\n%s' % res)
plot_decision_boundary(res.x, x, y)
pl.show()
print('============ Part 4: Optimizing fmin ====================')
res = op.fmin(cost_function_without_grad, init_theta, args=(x, y))
# print('Cost at theta found by fmin: %s' % cost)
print('Result by fmin:\n%s' % res)
plot_decision_boundary(res, x, y)
pl.show()
sp.seterrcall(saved_handler)
sp.seterr(**saved_err)
def __init__(self, d):
Layer.__init__(self, 1, d)
sigma = s.sqrt(1.0/d)
self.w = s.asmatrix(s.random.normal(0.0, sigma, (1, d)))
self.b = s.random.normal(0.0, sigma, 1)
def test():
'''a = mx('1,2,3;0,4,5;9,0,8')
print a.shape
print a.I
print mx.A'''
a = mx('1,2;3,2')
b = mx('1,0,0;0,1,1')
c = mx('1,0;0,1;1,0')
# print c*(a*b)
print a.shape[0]
ai = sp.identity(a.shape[1])
# aif = ai.flat
ail = ai.tolist()
newit = ail[0]
ail.append(newit)
print ail #.repeat(2,1)#.reshape((2,))
ailm = sp.asmatrix(ail)
print ailm
# c = mx('1,2;0,4;9,2')
# d = mx('1,2,0;4,9,2')
# print a*b
#每加入一个节点，得一A同型+1单位阵，，修改Isi+=1，与A相乘得新输出矩阵
linec = 0
def run():
input_layer_size = 400
hidden_layer_size = 25
num_labels = 10
print('Loading and Visualizing Data...')
X, y = digit_data['X'], digit_data['y']
m = X.shape[0]
# Randomly select 100 data points to display
sel = sp.random.permutation(m)
multiclass.displayData(X[sel, :])
print('Program paused. Press enter to continue.')
raw_input()
Theta1, Theta2 = weights_data['Theta1'], weights_data['Theta2']
neurnet.predict(Theta1, Theta2, X)
print('Program paused. Press enter to continue.')
raw_input()
rp = sp.random.permutation(m)
for i in range(0, m):
pred = neurnet.predict(Theta1, Theta2, sp.asmatrix(X[rp[i], :]))
print("Neural Network Prediction: %d (digit %d)\n" % (pred[0], sp.mod(pred[0], 10)))
def costFunctionReg(flattendTheta, X, y, lmbda):
"""
Calculate the cost and gradient for logistic regression
using regularization (helps with preventing overfitting
with many features)
"""
# numpy fmin function only allows flattened arrays instead of
# matrixes which is stupid so it has to be converted every time
flattendTheta = sp.asmatrix(flattendTheta)
(a, b) = flattendTheta.shape
if a < b:
theta = flattendTheta.T
else:
theta = flattendTheta
m = sp.shape(y)[0]
(J, grad) = costFunction(theta, X, y)
# f is a filter vector that will disregard regularization for theta0
f = sp.ones((theta.shape[0], 1))
f[0, 0] = 0
thetaFiltered = sp.multiply(theta, f)
J = J + (lmbda/(2.0 * m)) * (thetaFiltered.T.dot(thetaFiltered))
grad = grad + ((lmbda/m) * thetaFiltered).T
return (J, grad)
def run():
theta = sp.zeros((3, 1))
data = sp.copy(admission_data)
X = data[:, [0, 1]]
y = data[:, [2]]
m = sp.shape(y)[0]
# Add intercept term to x
X = sp.concatenate((sp.ones((m, 1)), X), axis=1)
"""
Part 1: Plotting
"""
print('Plotting data with + indicating (y = 1) examples and o indicating (y = 0) examples.')
logres.plotData(data)
plt.xlabel('Exam 1 score')
plt.ylabel('Exam 2 score')
plt.legend('Admitted', 'Not admitted')
plt.show()
print('Program paused. Press enter to continue.')
raw_input()
"""
Part 2: Compute Cost and Gradient
"""
(m, n) = X.shape
initial_theta = sp.zeros((n, 1))
(cost, grad) = logres.costFunction(initial_theta, X, y)
print('Cost at initial theta (zeros): ', cost)
print('Gradient at initial theta (zeros): ', grad)
print('Program paused. Press enter to continue.')
raw_input()
"""
Part 3: Optimizing using fminunc
"""
(theta, cost) = logres.find_minimum_theta(theta, X, y)
print('Cost at theta found by fmin: ', cost)
print('Theta: ', theta)
logres.plotDecisionBoundary(data, X, theta)
plt.show()
"""
Part 4: Predict and Accuracies
"""
prob = logres.sigmoid(sp.asmatrix([1, 45, 85]).dot(theta))
print('For a student with scores 45 and 85, we predict an admission probability of ', prob[0, 0])
print('Program paused. Press enter to continue.')
def __init__(self, neurons_num, d):
Layer.__init__(self, neurons_num, d)
links = 2 * self.h
sigma = s.sqrt(1.0/d)
self.w = s.asmatrix(s.random.normal(0.0, sigma, (links, d)))
self.b = s.random.normal(0.0, sigma, links)
def __init__(self, nodes, edges):
degree = dict(Counter([n for e in edges for n in e]))
self.vol = sum(degree)
self.n = N = len(nodes) # dimension of all matrix
self.A = sp.asmatrix(sp.zeros([N, N]))
self.D = sp.asmatrix(sp.zeros([N, N]))
for edge in edges: # fill affinity matrix
i = nodes.index(edge[0]) # source
j = nodes.index(edge[1]) # target
self.A[i, j] = self.A[j, i] = 1.0
for node in nodes: # fill degree matrix
i = nodes.index(node)
self.D[i, i] = degree[node]
self.L = sp.asmatrix(self.D - self.A)
def __init__(self, nodes, edges):
degree = dict(Counter([n for e in edges for n in e]))
self.vol = sum(degree)
self.n = N = len(nodes) # dimension of all matrix
self.A = sp.asmatrix(sp.zeros([N, N]))
self.D = sp.asmatrix(sp.zeros([N, N]))
for edge in edges: # fill affinity matrix
i = nodes.index(edge[0]) # source
j = nodes.index(edge[1]) # target
self.A[i, j] = self.A[j, i] = 1.0
for node in nodes: # fill degree matrix
i = nodes.index(node)
self.D[i, i] = degree[node]
self.L = sp.asmatrix(self.D - self.A)
def _generateRateMatrix(states, transitions, rates):
n_states = len(states)
Q = asmatrix(zeros((n_states, n_states)))
for (src, transition, dst) in transitions:
Q[src,dst] = rates[transition]
for i in xrange(n_states):
row = Q[i,:]
Q[i,i] = -sum(row)
return Q
def setVauleofMatrix(self,preoutputmatrix,setpositionlist,valuelist=None):
netDimNow = preoutputmatrix.shape[0]
poslen = mx.max(sp.asmatrix(setpositionlist))#sp.maximum(setpositionlist)+1#len(setpositionlist)
outmat = self.addrowcol_matrix(preoutputmatrix,poslen-netDimNow+1,poslen-netDimNow+1)
# print outmat.shape,
for pos in setpositionlist:
if pos:
outmat.itemset(pos,outmat.item(pos)+1)
return outmat
def cols_array2(my_m, n_iter):
my_m = asmatrix(my_m) # not strictly needed, but should be noop
# the following is valid as long as we assume a fixed matrix size (and not in a general function)
#v1 = asmatrix(full( (1, my_m.shape[0]), 1.0/my_m.shape[0]))
v1 = full((1, my_m.shape[0]), 1.0 / my_m.shape[0])
for i in xrange(n_iter):
#v1 = asmatrix(full( (1, my_m.shape[0]), 1.0/my_m.shape[0]))
#m1 = v1 * my_m
m1 = np.dot(v1, my_m)
m1[0] = 0 # just to avoid complaints
def pPca(data, dim):
"""Return a matrix which contains the first `dim` dimensions principal
components of data.
data is a matrix which's rows correspond to datapoints. Implementation of
the 'probabilistic PCA' algorithm.
"""
num = data.shape[1]
data = asmatrix(makeCentered(data))
# Pick a random reduction
W = asmatrix(standard_normal((num, dim)))
# Save for convergence check
W_ = W[:]
while True:
E = inv(W.T * W) * W.T * data.T
W, W_ = data.T * E.T * inv(E * E.T), W
if abs(W - W_).max() < 0.001:
break
return W.T
def pPca(data, dim):
"""Return a matrix which contains the first `dim` dimensions principal
components of data.
data is a matrix which's rows correspond to datapoints. Implementation of
the 'probabilistic PCA' algorithm.
"""
num = data.shape[1]
data = asmatrix(makeCentered(data))
# Pick a random reduction
W = asmatrix(standard_normal((num, dim)))
# Save for convergence check
W_ = W[:]
while True:
E = inv(W.T * W) * W.T * data.T
W, W_ = data.T * E.T * inv(E * E.T), W
if abs(W - W_).max() < 0.001:
break
return W.T
def addrow_mat(a, rows2addcnt=1):
ai = a #sp.identity(a.shape[1])
# aif = ai.flat
ail = ai.tolist()
for i in range(rows2addcnt):
rowtoadd = [0 for i in range(1, len(ail[0]) + 1)]
ail.append(rowtoadd)
ailm = sp.asmatrix(ail)
return ailm
def reduceDim(data, dim, func='pca'):
"""Reduce the dimension of datapoints to dim via principal component
analysis.
A matrix of shape (n, d) specifies n points of dimension d.
"""
try:
pcaFunc = globals()[func]
except KeyError:
raise ValueError('Unknown function to calc principal components')
pc = pcaFunc(data, dim)
return (pc * asmatrix(makeCentered(data)).T).T
def reduceDim(data, dim, func='pca'):
"""Reduce the dimension of datapoints to dim via principal component
analysis.
A matrix of shape (n, d) specifies n points of dimension d.
"""
try:
pcaFunc = globals()[func]
except KeyError:
raise ValueError('Unknown function to calc principal components')
pc = pcaFunc(data, dim)
return (pc * asmatrix(makeCentered(data)).T).T
def getoutMatrix(personcnt, ma_fans, cbm_frs, inv_mention, act_micorcnt,
worksfolder):
preoutmat = mx('1')
for i in range(1, personcnt + 1):
# print '-------------------',i
input = zip(
*[ma_fans[:i], cbm_frs[:i], inv_mention[:i], act_micorcnt[:i]])
inputm = sp.asmatrix(input, long)
min2mout_pat_mat = min2mout_mat(inputm,
preoutmat,
supposeMatrixDim=personcnt)
preoutmat = min2mout_pat_mat
return preoutmat
def compute_weighted_operator(fwd, inv, source_identities):
"""Function for computing a fidelity-weighted inverse operator.
Input arguments:
================
fwd : ndarray
The forward operator.
inv : ndarray
The original inverse operator.
source_identities : ndarray
Vector mapping sources to parcels or labels.
Output argument:
================
weighted_inv : ndarray
The fidelity-weighted inverse operator.
"""
"""Maybe one should test if unique non-negative values == max+1. This
is expected in the code."""
n_parcels = max(source_identities) + 1
"""Samples. Peaks at about 20 GB ram with 30 000 samples. Using too few
samples will give poor results."""
time_output = 6000
"""Samples to remove from ends to get rid of border effects."""
time_cut = 20
"""Original values 1, 31. Higher number wider span."""
widths = scipy.arange(5, 6)
"""Make and clone parcel time series to source time series."""
fwd, inv = asmatrix(fwd), asmatrix(inv)
parcel_series = make_series(n_parcels, time_output, time_cut, widths)
source_series = parcel_series[source_identities]
source_series[source_identities < 0] = 0
"""Forward then inverse model source series."""
source_series = inv * (fwd * source_series)
weighted_inv = _compute_weights(source_series, parcel_series,
source_identities, inv)
return weighted_inv
def deserialize_weights(self, w):
"""Takes array w, reshapes it and stores it as internal weights and
bias parameters
Note: this is not a serialization of the whole class, this
reconstruction uses dimensions already stored in object
"""
assert(len(w) == self.get_weights_len())
(wrows, wcols) = self.w.shape
wtotal = wrows * wcols
self.w = s.asmatrix(w[:wtotal]).reshape(wrows, wcols)
self.b = w[wtotal:]
def plv(x, y, source_identities):
""" Function for computing the complex phase-locking value.
x : Source time series
y : Parcel time series
source_identities : ndarray [sources]
Expected ids for parcels are 0 to n-1, where n is number of parcels,
and -1 for sources that do not belong to any parcel.
"""
"""Change to amplitude 1, keep angle using Euler's formula."""
x = np.exp(1j * (asmatrix(np.angle(x))))
y = np.exp(1j * (asmatrix(np.angle(y))))
"""Get cPLV needed for flips and weighting."""
cplv = np.zeros(len(source_identities), dtype='complex')
for i, identity in enumerate(source_identities):
"""Compute cPLV only of parcel source pairs of sources that
belong to that parcel. One source belong to only one parcel."""
if (source_identities[i] >= 0):
cplv[i] = (np.sum(
(np.asarray(y[identity])) * np.conjugate(np.asarray(x[i]))))
cplv /= np.shape(x)[1]
return cplv
def logistic_regression():
"""
Predicts the probability that a student will be admitted
to a university based on how well he did on two exams
Params:
exam1: Integer score
exam2: Integer score
"""
exam1 = int(request.args.get('exam1'))
exam2 = int(request.args.get('exam2'))
prob = sigmoid(sp.asmatrix([1, exam1, exam2]).dot(theta))
return jsonify({
'probability_accepted': prob[0,0]
})
def get_relationship_matrix(data, year, relationship_definition, def_args):
array_data = []
for export_country in countries.countries:
row = []
for import_country in countries.countries:
if relationship_definition(data, year, export_country, import_country,
def_args) and export_country != import_country:
row.append(1)
else:
row.append(0)
array_data.append(row)
a = scipy.array(array_data)
b = scipy.asmatrix(a)
return b
def plv(x, y, identities):
"""Function for computing phase-locking values between x and y.
Output arguments:
=================
cplv : ndarray
Complex-valued phase-locking values.
"""
"""Change to amplitude 1, keep angle using Euler's formula."""
x = scipy.exp(1j * (asmatrix(scipy.angle(x))))
y = scipy.exp(1j * (asmatrix(scipy.angle(y))))
"""Get cPLV needed for flips and weighting."""
cplv = scipy.zeros(len(identities), dtype='complex')
for i, identity in enumerate(identities):
"""Compute cPLV only of parcel source pairs of sources that
belong to that parcel. One source belong to only one parcel."""
if (identities[i] >= 0):
cplv[i] = (scipy.sum((scipy.asarray(y[identity])) *
scipy.conjugate(scipy.asarray(x[i]))))
cplv /= np.shape(x)[1]
return cplv
def costFunction(flattendTheta, X, y):
"""Calculate the cost and gradient for logistic regression"""
# numpy fmin function only allows flattened arrays instead of
# matrixes which is stupid so it has to be converted every time
flattendTheta = sp.asmatrix(flattendTheta)
(a, b) = flattendTheta.shape
if a < b:
theta = flattendTheta.T
else:
theta = flattendTheta
m = sp.shape(y)[0]
J = (1.0/m) * ((-y).T.dot(sp.log(sigmoid(X.dot(theta)))) - \
(-y + 1).T.dot(sp.log(-sigmoid(X.dot(theta)) + 1)))
grad = (1.0/m) * (sigmoid(X.dot(theta)) - y).T.dot(X)
return (J, grad)
def covar(samples):
"""
Calculate the covariance matrix as used by Estimator.
This is not the same as the Octave/Matlab function cov(), but is instead
equal to Mean [ sample.H * sample ], where sample is a single sample.
I.E., it is actually the second moment matrix.
Parameters
----------
samples : K x Numel or Numel x 0 complex ndarray
Complex samples for each of Numel antennas sampled at K times.
Returns
-------
return : Numel x Numel complex ndarray
Second moment matrix for complex random vector samples. Used by
Estimator.
"""
samples = sp.asmatrix(samples)
return ( (samples.H * samples) / samples.shape[0] )
def covar(samples):
"""
Calculate the covariance matrix as used by Estimator.
This is not the same as the Octave/Matlab function cov(), but is instead
equal to Mean [ sample.H * sample ], where sample is a single sample.
I.E., it is actually the second moment matrix.
Parameters
----------
samples : K x Numel or Numel x 0 complex ndarray
Complex samples for each of Numel antennas sampled at K times.
Returns
-------
return : Numel x Numel complex ndarray
Second moment matrix for complex random vector samples. Used by
Estimator.
"""
samples = sp.asmatrix(samples)
return ((samples.H * samples) / samples.shape[0])
def isHermitian(M, tol):
MM = asmatrix(M)
return norm(MM - MM.H) < tol
# pos = sum(x > 0 for x in guidance_matrix)
print(guidance_matrix)
ward = Ward(n_clusters=6, n_components=2, connectivity=guidance_matrix)
predicts = ward.fit_predict(self.A)
print(predicts)
# print(circles.keys(), len(circles.itervalues())
if __name__ == "__main__":
sp.set_printoptions(precision=2, suppress=True)
nodes = [1, 2, 3, 4, 5, 6]
edges = [[1, 2], [1, 3], [2, 3], [3, 4], [4, 5], [4, 6], [5, 6]]
Q1 = sp.asmatrix(sp.identity(6))
Q = sp.mat([[1.0, 1.0, 1.0, 1.0, -1.0, -1.0],
[1.0, 1.0, 1.0, 1.0, -1.0, -1.0],
[1.0, 1.0, 1.0, 1.0, -1.0, -1.0],
[1.0, 1.0, 1.0, 1.0, -1.0, -1.0],
[-1.0, -1.0, -1.0, -1.0, 1.0, 1.0],
[-1.0, -1.0, -1.0, -1.0, 1.0, 1.0]])
c = Cluster(nodes, edges)
vec = c.spectral(2)
print(vec)
# vec_idc = c.csp(Q1, 0, 3)
'''
# Constrained spectral clustering
def csp(self, Q, beta, K):
"""Collection of functions for converting between magnitude systems."""
import scipy
_usno_to_sdss_matrix = scipy.array([
[1, 0, 0, 0, 0], #u
[0, 1.06, -0.06, 0, 0], #g
[0, 0, 1.035, -0.035, 0], #r
[0, 0, 0.041, 1.0 - 0.041, 0], #i
[0, 0, 0, -0.03, 1.03] #z
])
_sdss_to_usno_matrix = scipy.asarray(scipy.asmatrix(_usno_to_sdss_matrix).I)
_sdss_to_usno_offset = scipy.array([[0], [0.06 * 0.53], [0.035 * 0.21],
[0.041 * 0.21], [-0.03 * 0.09]])
def sdss_to_usno(sdss_ugriz):
"""
Return the estimated USNO 1m estimated magnitudes from SDSS 2.5m ones.
Args:
sdss_ugriz(5xN scipy.array): The values of the u, g, r, and z
magnitudes in the SDSS 2.5m system. Each magnitude is a column.
Returns:
5 x N scipy array:
The values of the u', g', r', i', and z' magnitudes in the USNO 1m
system in the same format as the input.
"""
import csv
# IMPORT PROVIDED NETWORK DATA FROM FILES
nodes_file = open("../Data/QMEE_Net_Mat_nodes.csv", 'rb')
edges_file = open("../Data/QMEE_Net_Mat_edges.csv", 'rb')
nodes = []
csvread = csv.reader(nodes_file)
csvread.next() # skip header row
for row in csvread:
nodes.append(tuple(row))
nodes_file.close()
edges = sc.array
csvread = csv.reader(edges_file)
AdjNames = list(csvread.next())
tmp = csvread.next()
tmp = map(int, tmp)
Adj = sc.asmatrix(tmp)
for row in csvread:
tmp = map(int, row)
Adj = sc.append(Adj, [tmp], axis=0)
edges_file.close()
###### PLOTTING #######
plt.close('all')
G = nx.Graph(Adj)
nx.draw_circular(G)
plt.show()
def generateMotion (self): self.vecs = self.pcs*scipy.asmatrix(self.weights).T self.vecs = self.vecs.T self.vecs = scipy.asarray(self.vecs)[0]
import scipy import numpy from scipy.sparse import csc_matrix a = scipy.array([[1, 2, 3], [4, 5, 6], [4, 5, 6]]) b = scipy.asmatrix(a) print "b" print b print "b.tolist()[1][1]" print b.tolist()[1][1] print "numpy.dot(b, b)" print numpy.dot(b, b) squared_matrix = csc_matrix(b) * csc_matrix(b) squared_matrix1 = csc_matrix(b).getrow(1).getcol(1) #print squared_matrix print "squared_matrix.tobsr()" print squared_matrix.tobsr() print "abc" print "squared_matrix.todense()" print squared_matrix.todense() print "squared_matrix.todense().tolist()[1][1]" print squared_matrix.todense().tolist()[1][1]
# pos = sum(x > 0 for x in guidance_matrix)
print(guidance_matrix)
ward = Ward(n_clusters=6, n_components=2, connectivity=guidance_matrix)
predicts = ward.fit_predict(self.A)
print(predicts)
# print(circles.keys(), len(circles.itervalues())
if __name__ == "__main__":
sp.set_printoptions(precision=2, suppress=True)
nodes = [1, 2, 3, 4, 5, 6]
edges = [[1, 2], [1, 3], [2, 3], [3, 4], [4, 5], [4, 6], [5, 6]]
Q1 = sp.asmatrix(sp.identity(6))
Q = sp.mat([[1.0, 1.0, 1.0, 1.0, -1.0, -1.0],
[1.0, 1.0, 1.0, 1.0, -1.0, -1.0],
[1.0, 1.0, 1.0, 1.0, -1.0, -1.0],
[1.0, 1.0, 1.0, 1.0, -1.0, -1.0],
[-1.0, -1.0, -1.0, -1.0, 1.0, 1.0],
[-1.0, -1.0, -1.0, -1.0, 1.0, 1.0]])
c = Cluster(nodes, edges)
vec = c.spectral(2)
print(vec)
# vec_idc = c.csp(Q1, 0, 3)
def find_pattern_rotated(PF, pattern, image, rescale = 1.0, rotate=(-60,61,120),
roi_center=None, roi_size=(41,41), plot=False):
#Get current time to determine runtime of search
start_time = time.time()
#Initialize values needed later on
result = []
vmax = 0.0
vmin = sp.Inf
#Set region of interest
if roi_center is None:
roi_center = sp.array(im.shape[:2])/2.0 - 0.5
roi = center_roi_around(roi_center*rescale, roi_size)
#Give user some feedback on what is happening
print("Rescaling image and target by scale={rescale}.\n"
" image {0}x{1} px to {2:.2f}x{3:.02f} px."
.format(image.shape[0], image.shape[1],
image.shape[0]*rescale, image.shape[1]*rescale, rescale=rescale), flush=True)
print("ROI: center={0}, {1}, in unscaled image.\n"
" height={2}, width={3} in scaled image"
.format(roi_center[0], roi_center[1], roi_size[0], roi_size[1]))
if rotate[2]>1:
print("Now correlating rotations from {0}º to {1}º in {2} steps:"
.format(*rotate))
else:
print("Rotation is kept constant at {0}°".format(rotate[0]))
# Create rescaled copies of image and pattern, determine center coordinates of both
pattern_scaled = transform.rescale(pattern, rescale)
image_scaled = transform.rescale(image, rescale)
PF.set_image(image_scaled)
cols_scaled, rows_scaled = pattern_scaled.shape[:2]
pattern_scaled_center = sp.array((rows_scaled, cols_scaled))/2. - 0.5
cols, rows = pattern.shape[:2]
pattern_center = sp.array((rows, cols))/2. - 0.5
# Launch PatternFinder for all rotations defined in function input
rotations = sp.linspace(*rotate)
for r in rotations:
# Calculate transformation matrix for rotation around center of scaled pattern
rotation_matrix = rotation_transform_center(pattern_scaled,r,center_xy=pattern_scaled_center)
# Launch Patternfinder
out, min_coords, value = PF.find(transform.warp(pattern_scaled,rotation_matrix), image=None, roi=roi)
# Collect Min and Max values for plotting later on
outmax = out.max()
outmin = out.min()
if outmax > vmax:
vmax = outmax
if outmin < vmin:
vmin = outmin
# undo the rescale for the coordinates
min_coords = min_coords.astype(sp.float64) / rescale
# create a list of results for all rotations
result.append([r, min_coords, value, out])
# Progress bar... kind of :)
print(".",end="", flush=True)
print("")
print("took {0} seconds.".format(time.time()-start_time))
#Select the best result from the result list and extract its parameters
best_param_set = result[sp.argmin([r[2] for r in result])]
best_angle = best_param_set[0] # The rotation angle is the 0-th element in result
best_coord = best_param_set[1] # The coordinates are in the 2-nd element
best_value = best_param_set[2] # The actual value is the 3-rd element
# Calculate transformation to transform image onto pattern
move_to_center = transform.AffineTransform(translation=-(best_coord)[::-1])
move_back = transform.AffineTransform(translation=(best_coord[::-1]))
rotation = transform.AffineTransform(rotation=-sp.deg2rad(best_angle))
translation = transform.AffineTransform(translation=sp.asmatrix((best_coord-pattern_center)[::-1]))
T = translation + move_to_center + rotation + move_back
#Create a plot showing error over angle
if plot and rotate[2] > 1:
fig, ax = plt.subplots(1)
ax.plot([a[0] for a in result], [a[2] for a in result])
ax.set_xlabel('Angle (rotation)')
ax.set_ylabel('difference image-target')
plt.show()
#Create heat plot of where target is in image
if plot == 'all':
n_rows = int(sp.sqrt(len(result)))
n_cols = int(sp.ceil(len(result)/n_rows))
fig, ax = plt.subplots(n_rows, n_cols, squeeze=False, figsize = (2 * n_cols, 2 * n_rows))
fig.tight_layout(rect=[0, 0.03, 1, 0.97])
fig.suptitle("Correlation map of where target is in image\n", size=16)
n = 0
for i in range(n_rows):
for j in range(n_cols):
ax[i,j].axis("off")
if n < len(result):
ax[i,j].imshow(result[n][3], interpolation="nearest", cmap='cubehelix', vmin=vmin, vmax=vmax)
ax[i,j].annotate('Angle:{0:.2f}; Value:{1:.2f}'
.format(result[n][0],result[n][2]),[0,0])
n += 1
plt.show()
return T, best_value
sourceTimeSeries[i] = parcelTimeSeries[
identity] # Clone parcel time series to source space.
checkSourceTimeSeries = scipy.real(sourceTimeSeries[:]) # For checking
########## Forward then inverse model source series
# sourceTimeSeries = inverseOperator*(forwardOperator * sourceTimeSeries) this didn't work
sourceTimeSeries = np.dot(inverseOperator,
np.dot(forwardOperator,
sourceTimeSeries)) # this works
########## Change to amplitude 1, keep angle using Euler's formula.
sourceTimeSeries = scipy.exp(1j *
(scipy.asmatrix(scipy.angle(sourceTimeSeries))))
parcelTimeSeries = scipy.exp(1j *
(scipy.asmatrix(scipy.angle(parcelTimeSeries))))
########## Get cPLV needed for flips and weighting
cPLVArray = 1j * scipy.zeros(len(sourceIdentities),
dtype=float) # Initialize as zeros (complex).
for i, identity in enumerate(
sourceIdentities
): # Compute cPLV only of parcel source pairs of sources that belong to that parcel. One source belong to only one parcel.
if sourceIdentities[
i] >= 0: # Don't compute negative values. These should be sources not belonging to any parcel.
cPLVArray[i] = scipy.sum(
(scipy.asarray(parcelTimeSeries[identity])) *