def randomRotation():
"""
Get random rotation matrix.
Written by Michael Habeck.
:return: 3 x 3 array of float
:rtype: array
"""
alpha = R.random_sample() * 2 * N0.pi
gamma = R.random_sample() * 2 * N0.pi
beta = N0.arccos(2*(R.random_sample() - 0.5))
return eulerRotation(alpha, beta, gamma)
def randomRotation():
"""
Get random rotation matrix.
Written by Michael Habeck.
:return: 3 x 3 array of float
:rtype: array
"""
alpha = R.random_sample() * 2 * N0.pi
gamma = R.random_sample() * 2 * N0.pi
beta = N0.arccos(2 * (R.random_sample() - 0.5))
return eulerRotation(alpha, beta, gamma)
def test_change(dset_key, conn, fun):
"""Takes in a data set key and computes g(r) with different amounts of noise added
to the positions returns a list of the g(r)"""
max_rng = 100
nbins = 1000
buff = 1
cull_rate = 0.5
(fname, comp_num) = conn.execute(
"select fout,comp_key from comps where dset_key =?" + " and function = 'Iden'", (dset_key,)
).fetchone()
F = h5py.File(fname, "r")
frame = 10
x = F["/frame%(#)06d" % {"#": frame} + "/x_%(#)07d" % {"#": comp_num}]
y = F["/frame%(#)06d" % {"#": frame} + "/y_%(#)07d" % {"#": comp_num}]
parts = [a for a in itertools.izip(x, y) if nrm.random_sample() < cull_rate]
steps = 5
gs = []
for j in range(0, steps):
hc = sc.hash_case((np.ceil(np.max(x)), np.ceil(np.max(y))), max_rng)
gofr = sc.gofr_comp(nbins, max_rng)
for p in parts:
hc.add_particle(fun(p, j))
pass
hc.compute_corr(buff, gofr.add_particle)
gs.append(cord_pairs(gofr.bin_edges, gofr.vals))
return gs
def generate(self):
matrix = 8 * random_sample((3, 8)) - 4
for i in range(3):
for j in range(8):
if i + 6 == j:
matrix[i][j] = 0
return matrix
def random_Butter(n=(5, 10),
Wc=(0.1, 0.8),
W1=(0.1, 0.5),
W2=(0.5, 0.8),
form=None,
onlyEven=True,
seed=None):
"""
Generate a n-th order Butterworh filters
Parameters
----------
- n: (int) The order of the filter
- Wc: used if btype is 'lowpass' or 'highpass'
Wc is a tuple (min,max) for the cut frequency
- W1 and W2: used if btype is ‘bandpass’, ‘bandstop’
W1 and W2 are tuple (min,max) for the two start/stop frequencies
- form: (string) {None, ‘lowpass’, ‘highpass’, ‘bandpass’, ‘bandstop’}. Gives the type of filter. If None, the type is randomized
- onlyEven: if True, only even order filter are generated
- seed: if not None, indicates the seed toi use for the random part (in order to be reproductible, the seed is stored in the name of the filter)
"""
# change the seed if asked
if seed:
numpy_seed(seed)
# choose a form if asked
if form is None:
form = choice(("lowpass", "highpass", "bandpass", "bandstop"))
# choose Wn
if form in ("bandpass", "bandstop"):
# choose 2 frequencies
if W2[1] <= W1[0]:
raise ValueError("iter_random_Butter: W1 should be lower than W2")
Wn1 = (W1[1] - W1[0]) * random_sample() + W1[0]
Wn2 = (W2[1] - W2[0]) * random_sample() + W2[0]
while Wn2 <= Wn1:
Wn2 = (W2[1] - W2[0]) * random_sample() + W2[0]
W = [Wn1, Wn2]
else:
# choose 1 frequency
W = (Wc[1] - Wc[0]) * random_sample() + Wc[0]
# choose order
order = randint(*n)
if onlyEven and order % 2 == 0:
order += 1
return Butter(order, W, form, name='Butterworth-random-%d' % seed)
def random_TF(n=(5, 10), Wc=(0.1, 0.8), W1=(0.1, 0.5), W2=(0.5, 0.8)):
"""Generate one n-th order stable butterworth filter ((num, den) of the transfer function)"""
# choose a form
form = choice(['lowpass', 'highpass', 'bandpass', 'bandstop'])
# choose Wn
if form in ("bandpass", "bandstop"):
# choose 2 frequencies such that Wn2<=Wn1
Wn1 = (W1[1] - W1[0]) * random_sample() + W1[0]
Wn2 = (W2[1] - W2[0]) * random_sample() + W2[0]
while Wn2 <= Wn1:
Wn2 = (W2[1] - W2[0]) * random_sample() + W2[0]
W = [Wn1, Wn2]
else:
# choose 1 frequency
W = (Wc[1] - Wc[0]) * random_sample() + Wc[0]
# choose order
order = randint(*n)
num, den = butter(order, W, form)
return num, den
def test_FuzzyCluster( self):
"""FuzzyCluster test"""
import gnuplot as G
x1 = R.random_sample((500,2))
x2 = R.random_sample((500,2)) + 1
x3 = R.random_sample((500,2)) + 2
self.x = N0.concatenate((x1, x2, x3))
self.fuzzy = FuzzyCluster(self.x, n_cluster=5, weight=1.5)
self.centers = self.fuzzy.go(1.e-30, n_iterations=50, nstep=10,
verbose=self.local)
if self.local:
print "cluster centers are displayed in green"
G.scatter( self.x, self.centers )
self.assertEqual( N0.shape(self.centers), (5, 2) )
def test_FuzzyCluster( self):
"""FuzzyCluster test"""
import biskit.gnuplot as G
x1 = R.random_sample((500,2))
x2 = R.random_sample((500,2)) + 1
x3 = R.random_sample((500,2)) + 2
self.x = N0.concatenate((x1, x2, x3))
self.fuzzy = FuzzyCluster(self.x, n_cluster=5, weight=1.5)
self.centers = self.fuzzy.go(1.e-30, n_iterations=50, nstep=10,
verbose=self.local)
if self.local:
print("cluster centers are displayed in green")
G.scatter( self.x, self.centers )
self.assertEqual( N0.shape(self.centers), (5, 2) )
def random_Elliptic(n=(5, 10),
rp=(10, 50),
rs=(10, 50),
Wc=(0.1, 0.8),
W1=(0.1, 0.5),
W2=(0.5, 0.8),
form=None,
seed=None,
quant=None):
"""
Generate a n-th order Butterworh filters
Parameters
----------
- n: (int) The order of the filter
- Wc: used if btype is 'lowpass' or 'highpass'
Wc is a tuple (min,max) for the cut frequency
- W1 and W2: used if btype is ‘bandpass’, ‘bandstop’
W1 and W2 are tuple (min,max) for the two start/stop frequencies
- form: (string) {None, ‘lowpass’, ‘highpass’, ‘bandpass’, ‘bandstop’}. Gives the type of filter. If None, the type is randomized
- quant: quantized the coefficients with quant bits (None by default -> no quantization)
- seed: if not None, indicates the seed toi use for the random part (in order to be reproductible, the seed is stored in the name of the filter)
"""
# change the seed if asked
if seed:
numpy_seed(seed)
# choose a form if asked
if form is None:
form = choice(("lowpass", "highpass", "bandpass", "bandstop"))
# choose Wn
if form in ("bandpass", "bandstop"):
# choose 2 frequencies
if W2[1] <= W1[0]:
raise ValueError(
"iter_random_Elliptic: W1 should be lower than W2")
Wn1 = (W1[1] - W1[0]) * random_sample() + W1[0]
Wn2 = (W2[1] - W2[0]) * random_sample() + W2[0]
while Wn2 <= Wn1:
Wn2 = (W2[1] - W2[0]) * random_sample() + W2[0]
W = [Wn1, Wn2]
else:
# choose 1 frequency
W = (Wc[1] - Wc[0]) * random_sample() + Wc[0]
# choose rp and rs
rip = (rp[1] - rp[0]) * random_sample() + rp[0]
rst = (rs[1] - rs[0]) * random_sample() + rs[0]
# choose order
order = randint(*n)
# do we quantified the parameter (to be able to print them exactly, for example)
if quant:
rip = quantify(rip, quant)
rst = quantify(rst, quant)
if isinstance(W, list):
W = [quantify(W[0], quant), quantify(W[1], quant)]
return Elliptic(order,
rp=rip,
rs=rst,
Wn=W,
etype=form,
name='Elliptic-random-%d' % seed)
def __random_translation( self ):
"""
Random translation on a sphere around 0,0,0 with fixed radius
The radius is the sum of the (max) radius of receptor and ligand
@return: translation array 3 x 1 of float
@rtype: array
"""
radius = (self.d_max_rec + self.d_max_lig) / 2.0
xyz = R.random_sample( 3 ) - 0.5
scale = radius*1.0 / N0.sqrt( N0.sum( xyz**2 ) )
return scale * xyz
def __random_translation( self ):
"""
Random translation on a sphere around 0,0,0 with fixed radius
The radius is the sum of the (max) radius of receptor and ligand
@return: translation array 3 x 1 of float
@rtype: array
"""
radius = (self.d_max_rec + self.d_max_lig) / 2.0
xyz = R.random_sample( 3 ) - 0.5
scale = radius*1.0 / N0.sqrt( N0.sum( xyz**2 ) )
return scale * xyz
def create_membership_matrix(self):
"""
Create a random membership matrix.
@return: random array of shape length of data to
cluster times number of clusters
@rtype: array('f')
"""
## default signature has changed oldnumeric->numpy
if (self.seedx == 0 or self.seedy == 0):
R.seed()
else:
R.seed((self.seedx, self.seedy))
r = R.random_sample((self.npoints, self.n_cluster))
return N0.transpose(r / N0.sum(r))
def create_membership_matrix(self):
"""
Create a random membership matrix.
@return: random array of shape length of data to
cluster times number of clusters
@rtype: array('f')
"""
## default signature has changed oldnumeric->numpy
if (self.seedx==0 or self.seedy==0):
R.seed()
else:
R.seed((self.seedx, self.seedy))
r = R.random_sample((self.npoints, self.n_cluster))
return N0.transpose(r / N0.sum(r))
def compute_clusters(self, pXY, qXhatYhat, qXhat, qXxhat, cXY, axis):
""" Compute the best cluster assignment along a single axis, given all the distributions and clusters on other axes.
Args:
pXY: the original probability distribution matrix
qXhatYhat: the joint distribution over the clusters
qXhat: the marginal distributions of qXhatYhat
qXxhat: the distribution conditioned on the clustering in a list
cXY: current cluster assignments along each dimension
axis: the axis (dimension) over which clusters are being computed
Return:
Best cluster assignment along a single axis as a list
"""
if not isinstance(pXY, SparseMatrix) or not isinstance(qXhatYhat, SparseMatrix):
raise Exception("Arguments to compute_clusters not an instance of SparseMatrix.")
# To assign clusters, we calculate argmin_xhat D(p(Y,Z|x) || q(Y,Z|xhat)),
# where D(P|Q) = \sum_i P_i log (P_i / Q_i)
dPQ = np.zeros(shape=(pXY.N[axis], qXhatYhat.N[axis]))
# iterate though all non-zero elements; here we are making use of the sparsity to reduce computation
for coords, p_i in pXY.nonzero_elements.iteritems():
coord_this_axis = coords[axis]
px = self.pX[axis][coord_this_axis]
p_i = 1 if px == 0 else p_i / px # calculate p(y|x) = p(x,y)/p(x), but we should be careful if px == 0
current_cluster_assignments = [cXY[i][coords[i]] for i in
xrange(self.dim)] # cluster assignments on each axis
for xhat in xrange(self.K[axis]):
current_cluster_assignments[axis] = xhat # temporarily assign dth dimension to this xhat
current_qXhatYhat = qXhatYhat.get(tuple(current_cluster_assignments))
current_qXhat = qXhat[axis][xhat]
q_i = 1.0
if current_qXhatYhat == 0 and current_qXhat == 0:
q_i = 0 # Here we define 0/0=0
else:
q_i *= current_qXhatYhat / current_qXhat
for i in xrange(self.dim):
if i == axis: continue
q_i *= qXxhat[i][coords[i]]
if q_i == 0: # this can definitely happen if cluster joint distribution has zero element
dPQ[coord_this_axis, xhat] = INFINITE
else:
dPQ[coord_this_axis, xhat] += p_i * math.log(p_i / q_i)
# add random jitter to break ties
dPQ += self.jitter_max * random_sample(dPQ.shape)
return list(dPQ.argmin(1)) # return the closest cluster assignment under KL-divergence
def solve(self):
self.t = 0
self.initialize()
self.evaluate()
best = np.max(self.fitness)
newpop1 = self.population
bestIndex = np.argmax(self.fitness)
self.best = copy.deepcopy(self.population[bestIndex])
while (self.t < self.MAXGEN and best != 32):
self.t += 1
newpop = []
arr = []
fitness_sum = []
population_sum = []
for i in range(self.sizepop):
bestmatrix = self.best.matrix
bestmatrix1 = 8 * random_sample((3, 8)) - 4
ind = GAIndividual()
ind.matrix = bestmatrix1
newpop.append(ind)
population_sum = newpop1 + newpop
for i in range(2 * self.sizepop):
fitness_sum.append(population_sum[i].calculateFitness())
fitness_sum = np.array(fitness_sum)
arr = np.argsort(-fitness_sum)
for i in range(self.sizepop):
newpop[i] = population_sum[arr[i]]
self.population = []
self.population = newpop
self.evaluate()
best = np.max(self.fitness)
bestIndex = np.argmax(self.fitness)
best = np.max(self.fitness)
newpop1 = self.population
if best == 32:
break
matrix2 = self.population[bestIndex].matrix
for i in range(3):
for j in range(8):
print(matrix2[i][j], end=' ')
print("")
print(self.iteration)
def initialize_cluster_centers(self, pXY, K):
""" Initializes the cluster assignments along each axis, by first selecting k centers,
and then map each row to its closet center under cosine similarity.
Args:
pXY: original data matrix
K: numbers of clusters desired in each dimension
Return:
new_C: a list of list of cluster id that the current index in the current axis is assigned to.
"""
if not isinstance(pXY, SparseMatrix):
raise Exception("Matrix argument to initialize_cluster_centers is not an instance of SparseMatrix.")
new_C = [[-1] * Ni for Ni in pXY.N]
for axis in xrange(len(K)): # loop over each dimension
# choose cluster centers
axis_length = pXY.N[axis]
center_indices = random.sample(xrange(axis_length), K[axis])
cluster_ids = {}
for i in xrange(K[axis]): # assign identifiers to clusters
center_index = center_indices[i]
cluster_ids[center_index] = i
centers = defaultdict(lambda: defaultdict(float)) # all nonzero indices for each center
for coords in pXY.nonzero_elements:
coord_this_axis = coords[axis]
if coord_this_axis in cluster_ids: # is a center
reduced_coords = tuple([coords[i] for i in xrange(len(coords)) if i != axis]) # coords without the current axis
centers[cluster_ids[coord_this_axis]][reduced_coords] = pXY.nonzero_elements[coords] # (cluster_id, other coords) -> value
# assign rows to clusters
scores = np.zeros(shape=(pXY.N[axis], K[axis])) # scores: axis_size x cluster_number
denoms_P = np.zeros(shape=(pXY.N[axis]))
denoms_Q = np.zeros(shape=(K[axis]))
for coords in pXY.nonzero_elements:
coord_this_axis = coords[axis]
if coord_this_axis in center_indices:
continue # don't reassign cluster centers, please
reduced_coords = tuple([coords[i] for i in xrange(len(coords)) if i != axis])
for cluster_index in cluster_ids:
xhat = cluster_ids[cluster_index] # need cluster ID, not the axis index
if reduced_coords in centers[xhat]: # overlapping point
P_i = pXY.nonzero_elements[coords]
Q_i = centers[xhat][reduced_coords]
scores[coords[axis]][xhat] += P_i * Q_i # now doing based on cosine similarity
denoms_P[coords[axis]] += P_i * P_i # magnitude of this slice of original matrix
denoms_Q[xhat] += Q_i * Q_i # magnitude of cluster centers
# normalize scores
scores = divide(scores, outer(sqrt(denoms_P), sqrt(denoms_Q)))
scores[scores == 0] = -1.0
# add random jitter to scores to handle tie-breaking
scores += self.jitter_max * random_sample(scores.shape)
new_cXYi = list(scores.argmax(1)) # this needs to be argmax because cosine similarity
# make sure to assign the cluster centers to themselves
for center_index in cluster_ids:
new_cXYi[center_index] = cluster_ids[center_index]
# ensure numbers of clusters are correct
self.ensure_correct_number_clusters(new_cXYi, K[axis])
new_C[axis] = new_cXYi
return new_C
std_deviation = 0.3
noise = std_deviation * randn(N)
t = tlab + noise
for idx, val in enumerate(mDegrees):
w = polyfit(x, t, val)
row = idx / 2
col = idx % 2
fig = ax[row, col]
fig.set_title('M=%d' % val)
fig.plot(xlab, ylab)
fig.plot(x, t, 'ro')
fig.plot(xlab, polynomial(w, xlab), 'g')
NTest = 8
xTest = random_sample(NTest)
yTest = sin(2 * pi * xTest) + randn(NTest) * std_deviation
test_err = []
train_err = []
maxOrder = 10
figError = plt.figure()
for m in range(0, maxOrder):
weights = polyfit(x, t, m)
train_err.append(rms(weights, x, t))
test_err.append(rms(weights, xTest, yTest))
plt.xlabel("M")
def random(shape=[]):
"random(n) or random([n, m, ...]) returns array of random numbers"
if shape == []:
shape = None
return mt.random_sample(shape)
def random(shape=[]):
"random(n) or random([n, m, ...]) returns array of random numbers"
if shape == []:
shape = None
return mt.random_sample(shape)
t = tlab + noise
for idx, val in enumerate(mDegrees):
w = polyfit(x, t, val)
row = idx / 2
col = idx % 2
fig = ax[row, col]
fig.set_title('M=%d' % val)
fig.plot(xlab, ylab)
fig.plot(x, t, 'ro')
fig.plot(xlab, polynomial(w, xlab), 'g')
NTest = 8
xTest = random_sample(NTest)
yTest = sin(2 * pi * xTest) + randn(NTest) * std_deviation
test_err = []
train_err = []
maxOrder = 10
figError = plt.figure()
for m in range(0, maxOrder):
weights = polyfit(x, t, m)
train_err.append(rms(weights, x, t))
test_err.append(rms(weights, xTest, yTest))
