def __init__(self, ax, Y, clusters, color=None, pmax=0.05, ptrend=0.1,
xdim='time', title=None):
uts_args = _base.find_uts_args(Y, False, color)
self._bottom, self._top = _base.find_vlim_args(Y)
if title:
if '{name}' in title:
title = title.format(name=Y.name)
ax.set_title(title)
_plt_uts(ax, Y, xdim=xdim, **uts_args)
if np.any(Y.x < 0) and np.any(Y.x > 0):
ax.axhline(0, color='k')
# pmap
self.cluster_plt = _plt_uts_clusters(ax, clusters, pmax, ptrend, color)
# save ax attr
self.ax = ax
x = Y.get_dim(xdim).x
self.xlim = (x[0], x[-1])
ax.set_xlim(*self.xlim)
ax.set_ylim(bottom=self._bottom, top=self._top)
def _validate_covars(covars, cvtype, nmix, n_dim):
from scipy import linalg
if cvtype == 'spherical':
if len(covars) != nmix:
raise ValueError("'spherical' covars must have length nmix")
elif np.any(covars <= 0):
raise ValueError("'spherical' covars must be non-negative")
elif cvtype == 'tied':
if covars.shape != (n_dim, n_dim):
raise ValueError("'tied' covars must have shape (n_dim, n_dim)")
elif (not np.allclose(covars, covars.T)
or np.any(linalg.eigvalsh(covars) <= 0)):
raise ValueError("'tied' covars must be symmetric, "
"positive-definite")
elif cvtype == 'diag':
if covars.shape != (nmix, n_dim):
raise ValueError("'diag' covars must have shape (nmix, n_dim)")
elif np.any(covars <= 0):
raise ValueError("'diag' covars must be non-negative")
elif cvtype == 'full':
if covars.shape != (nmix, n_dim, n_dim):
raise ValueError("'full' covars must have shape "
"(nmix, n_dim, n_dim)")
for n, cv in enumerate(covars):
if (not np.allclose(cv, cv.T)
or np.any(linalg.eigvalsh(cv) <= 0)):
raise ValueError("component %d of 'full' covars must be "
"symmetric, positive-definite" % n)
def nanallclose(x, y, rtol=1.0e-5, atol=1.0e-8):
"""Numpy allclose function which allows NaN
Input
x, y: Either scalars or numpy arrays
Output
True or False
Returns True if all non-nan elements pass.
"""
xn = numpy.isnan(x)
yn = numpy.isnan(y)
if numpy.any(xn != yn):
# Presence of NaNs is not the same in x and y
return False
if numpy.all(xn):
# Everything is NaN.
# This will also take care of x and y being NaN scalars
return True
# Filter NaN's out
if numpy.any(xn):
x = x[-xn]
y = y[-yn]
# Compare non NaN's and return
return numpy.allclose(x, y, rtol=rtol, atol=atol)
def _testUniformSampleMultiDimensional(self):
# DISABLED: Please enable this test once b/issues/30149644 is resolved.
with self.test_session():
batch_size = 2
a_v = [3.0, 22.0]
b_v = [13.0, 35.0]
a = constant_op.constant([a_v] * batch_size)
b = constant_op.constant([b_v] * batch_size)
uniform = uniform_lib.Uniform(low=a, high=b)
n_v = 100000
n = constant_op.constant(n_v)
samples = uniform.sample(n)
self.assertEqual(samples.get_shape(), (n_v, batch_size, 2))
sample_values = self.evaluate(samples)
self.assertFalse(
np.any(sample_values[:, 0, 0] < a_v[0]) or
np.any(sample_values[:, 0, 0] >= b_v[0]))
self.assertFalse(
np.any(sample_values[:, 0, 1] < a_v[1]) or
np.any(sample_values[:, 0, 1] >= b_v[1]))
self.assertAllClose(
sample_values[:, 0, 0].mean(), (a_v[0] + b_v[0]) / 2, atol=1e-2)
self.assertAllClose(
sample_values[:, 0, 1].mean(), (a_v[1] + b_v[1]) / 2, atol=1e-2)
def tag_sites(self, scaled_positions, symprec=1e-3):
"""Returns an integer array of the same length as *scaled_positions*,
tagging all equivalent atoms with the same index.
Example:
>>> from ase.lattice.spacegroup import Spacegroup
>>> sg = Spacegroup(225) # fcc
>>> sg.tag_sites([[0.0, 0.0, 0.0],
... [0.5, 0.5, 0.0],
... [1.0, 0.0, 0.0],
... [0.5, 0.0, 0.0]])
array([0, 0, 0, 1])
"""
scaled = np.array(scaled_positions, ndmin=2)
scaled %= 1.0
scaled %= 1.0
tags = -np.ones((len(scaled), ), dtype=int)
mask = np.ones((len(scaled), ), dtype=np.bool)
rot, trans = self.get_op()
i = 0
while mask.any():
pos = scaled[mask][0]
sympos = np.dot(rot, pos) + trans
# Must be done twice, see the scaled_positions.py test
sympos %= 1.0
sympos %= 1.0
m = ~np.all(np.any(np.abs(scaled[np.newaxis,:,:] -
sympos[:,np.newaxis,:]) > symprec,
axis=2), axis=0)
assert not np.any((~mask) & m)
tags[m] = i
mask &= ~m
i += 1
return tags
def parseData(data, userType):
if not np.any(data.columns == "userid"):
sys.exit("'userid' column not found! Module = 'parseData'")
if not np.any(data.columns == "adid"):
sys.exit("'adid' column not found! Module = 'parseData'")
# Handle datatypes
try:
data["adid"] = data["adid"].astype(str)
print(" 'data' datatypes converted successfully.")
except:
sys.exit(" Could not convert 'data' datatypes: Exit.")
# Filter out rows that contain "" in the userid column
nbefore = data.shape[0]
# Only done for loginids
if userType == "login":
try:
data = data[data["userid"].apply(lambda x: x.isdigit())]
except:
print(" Warning: Failed to parse 'userid' column.")
try:
data = data.loc[data["userid"].astype(int) > 999999, :]
except:
print(" Warning: Failed to filter 'userid' column.")
# Check returned results
if data.shape[0] < 10:
sys.exit(" No data left after filtering. System exit.")
print(" Number of rows removed in 'data' = %i" % (data.shape[0] - nbefore))
return data
def testDtype(self):
with self.test_session():
d = array_ops.fill([2, 3], 12., name="fill")
self.assertEqual(d.get_shape(), [2, 3])
# Test default type for both constant size and dynamic size
z = array_ops.zeros([2, 3])
self.assertEqual(z.dtype, dtypes_lib.float32)
self.assertEqual([2, 3], z.get_shape())
self.assertAllEqual(z.eval(), np.zeros([2, 3]))
z = array_ops.zeros(array_ops.shape(d))
self.assertEqual(z.dtype, dtypes_lib.float32)
self.assertEqual([2, 3], z.get_shape())
self.assertAllEqual(z.eval(), np.zeros([2, 3]))
# Test explicit type control
for dtype in [
dtypes_lib.float32, dtypes_lib.float64, dtypes_lib.int32,
dtypes_lib.uint8, dtypes_lib.int16, dtypes_lib.int8,
dtypes_lib.complex64, dtypes_lib.complex128, dtypes_lib.int64,
dtypes_lib.bool, dtypes_lib.string
]:
z = array_ops.zeros([2, 3], dtype=dtype)
self.assertEqual(z.dtype, dtype)
self.assertEqual([2, 3], z.get_shape())
z_value = z.eval()
self.assertFalse(np.any(z_value))
self.assertEqual((2, 3), z_value.shape)
z = array_ops.zeros(array_ops.shape(d), dtype=dtype)
self.assertEqual(z.dtype, dtype)
self.assertEqual([2, 3], z.get_shape())
z_value = z.eval()
self.assertFalse(np.any(z_value))
self.assertEqual((2, 3), z_value.shape)
def compare(neurons1, spikes1, neurons2, spikes2):
import matplotlib.pyplot as plt
[sn1,ss1]=sort(neurons1,spikes1)
[sn2,ss2]=sort(neurons2,spikes2)
#sn1 = neurons1
#ss1 = spikes1
#sn2 = neurons2
#ss2 = spikes2
in1 = np.in1d(sn1, sn2)
in2 = np.in1d(sn2, sn1)
nin = len(sn1[in1])
print "Neuron in 1 but not in 2:", len(sn1)-nin
print "Neuron in 2 but not in 1:", len(sn2)-nin
for i in range(0,nin):
if np.any(ss1[in1][i] > 0) and np.any(ss2[in2][i] > 0):
if (len(ss1[in1][i]) == len(ss2[in2][i])):
plt.plot(ss1[in1][i]-ss2[in2][i], i*np.ones([len(ss1[in1][i]),1]), '*')
else:
print "For neuron", sn1[in1][i], "difference in length of spiketrains: ", len(ss1[in1][i]) - len(ss2[in2][i])
print "ss1:", ss1[in1][i]
print "ss2:", ss2[in2][i]
#plt.plot(ss1[in1][i][:np.min(len(ss1[in1][i]), len(ss2[in2][i]))]-ss2[in2][i][:np.min(len(ss1[in1][i]), len(ss2[in2][i]))], i*np.ones([np.min(len(ss1[in1][i]), len(ss2[in2][i])),1]), '*')
plt.show()
def predict(self, session, X, y=None):
"""Make predictions from the provided model."""
# If y is given, the loss is also calculated
# We deactivate dropout by setting it to 1
dp = 1
losses = []
results = []
if np.any(y):
data = data_iterator(X, y, batch_size=self.config.batch_size,
label_size=self.config.label_size, shuffle=False)
else:
data = data_iterator(X, batch_size=self.config.batch_size,
label_size=self.config.label_size, shuffle=False)
for step, (x, y) in enumerate(data):
feed = self.create_feed_dict(input_batch=x, dropout=dp)
if np.any(y):
feed[self.labels_placeholder] = y
loss, preds = session.run(
[self.loss, self.predictions], feed_dict=feed)
losses.append(loss)
else:
preds = session.run(self.predictions, feed_dict=feed)
predicted_indices = preds.argmax(axis=1)
results.extend(predicted_indices)
return np.mean(losses), results
def _LoadHeader(self):
ConfigLoader._LoadHeader(self)
bc = self.Domain.BlockCounts
nBlocks = self.Domain.TotalBlocks
# Number of blocks in its neighbourhood.
self.BlockNeighbourhoodSize = np.zeros(nBlocks, dtype=np.uint8)
# Number of blocks in the neighbourhood that are available
self.BlockNeighbourhoodAvailable = np.zeros(nBlocks, dtype=np.uint8)
# Number of blocks in the neighbourhood that are done
self.BlockNeighbourhoodDone = np.zeros(nBlocks, dtype=np.uint8)
# Is the block itself done
self.IsBlockDone = np.zeros(nBlocks, dtype=np.bool)
# Lock to ensure only one thread at a time updates IsBlockDone
# and BlockNeighbourhoodDone
self.DoneLock = threading.RLock()
# Compute the size of each block's neighbourhood
for bIjk, bIdx in self.Domain.BlockIndexer.IterBoth():
for i, delta in enumerate(self.NeighbourhoodOffsets):
nIdx = bIdx + delta
if np.any(nIdx < 0) or np.any(nIdx >= bc):
continue
nIjk = self.Domain.BlockIndexer.NdToOne(nIdx)
self.BlockNeighbourhoodSize[nIjk] += 1
continue
continue
return
def allowed_region( V_nj, ave_j ):
# read PCs
PC1 = V_nj[0]
PC2 = V_nj[1]
n_band = len( PC1 )
band_ticks = np.arange( n_band )
x_ticks = np.linspace(-0.4,0.2,RESOLUTION)
y_ticks = np.linspace(-0.2,0.4,RESOLUTION)
x_mesh, y_mesh, band_mesh = np.meshgrid( x_ticks, y_ticks, band_ticks, indexing='ij' )
vec_mesh = x_mesh * PC1[ band_mesh ] + y_mesh * PC2[ band_mesh ] + ave_j[ band_mesh ]
x_grid, y_grid = np.meshgrid( x_ticks, y_ticks, indexing='ij' )
prohibited_grid = np.zeros_like( x_grid )
for ii in xrange( len( x_ticks ) ) :
for jj in xrange( len( y_ticks ) ) :
if np.any( vec_mesh[ii][jj] < 0. ) :
prohibited_grid[ii][jj] = 1
if np.any( vec_mesh[ii][jj] > 1. ) :
prohibited_grid[ii][jj] = 3
elif np.any( vec_mesh[ii][jj] > 1. ) :
prohibited_grid[ii][jj] = 2
else :
prohibited_grid[ii][jj] = 0
return x_grid, y_grid, prohibited_grid
def __init__(self, x, y):
assert np.ndim(x)==2 and np.ndim(y)==2 and np.shape(x)==np.shape(y), \
'x and y must be 2D arrays of the same size.'
if np.any(np.isnan(x)) or np.any(np.isnan(y)):
x = np.ma.masked_where( (isnan(x)) | (isnan(y)) , x)
y = np.ma.masked_where( (isnan(x)) | (isnan(y)) , y)
self.x_vert = x
self.y_vert = y
mask_shape = tuple([n-1 for n in self.x_vert.shape])
self.mask_rho = np.ones(mask_shape, dtype='d')
# If maskedarray is given for verticies, modify the mask such that
# non-existant grid points are masked. A cell requires all four
# verticies to be defined as a water point.
if isinstance(self.x_vert, np.ma.MaskedArray):
mask = (self.x_vert.mask[:-1,:-1] | self.x_vert.mask[1:,:-1] | \
self.x_vert.mask[:-1,1:] | self.x_vert.mask[1:,1:])
self.mask_rho = np.asarray(~(~np.bool_(self.mask_rho) | mask), dtype='d')
if isinstance(self.y_vert, np.ma.MaskedArray):
mask = (self.y_vert.mask[:-1,:-1] | self.y_vert.mask[1:,:-1] | \
self.y_vert.mask[:-1,1:] | self.y_vert.mask[1:,1:])
self.mask_rho = np.asarray(~(~np.bool_(self.mask_rho) | mask), dtype='d')
self._calculate_subgrids()
self._calculate_metrics()
def fill_betweenx_discontinuous(ax, ymin, ymax, x, freq=1, **kwargs):
"""Fill betwwen x even if x is discontinuous clusters
Parameters
----------
ax : axis
x : list
Returns
-------
ax : axis
"""
x = np.array(x)
min_gap = 1.1 / freq
while np.any(x):
# If with single time point
if len(x) > 1:
xmax = np.where((x[1:] - x[:-1]) > min_gap)[0]
else:
xmax = [0]
# If continuous
if not np.any(xmax):
xmax = [len(x) - 1]
ax.fill_betweenx((ymin, ymax), x[0], x[xmax[0]], **kwargs)
# remove from list
x = x[(xmax[0] + 1) :]
return ax
def test_using_gpu_1(self):
# I'm checking if this compiles and runs
from theano import function, config, shared, sandbox
import theano.tensor as T
import numpy
import time
vlen = 10 * 30 * 70 # 10 x #cores x # threads per core
iters = 10
rng = numpy.random.RandomState(22)
x = shared(numpy.asarray(rng.rand(vlen), config.floatX))
f = function([], T.exp(x))
# print f.maker.fgraph.toposort()
t0 = time.time()
for i in xrange(iters):
r = f()
t1 = time.time()
print 'Looping %d times took' % iters, t1 - t0, 'seconds'
print 'Result is', r
if numpy.any([isinstance(x.op, T.Elemwise) for x in f.maker.fgraph.toposort()]):
print 'Used the cpu'
else:
print 'Used the gpu'
if theano.config.device.find('gpu') > -1:
assert not numpy.any( [isinstance(x.op,T.Elemwise) for x in f.maker.fgraph.toposort()])
else:
assert numpy.any([isinstance(x.op, T.Elemwise) for x in f.maker.fgraph.toposort()])
def test_using_gpu_3(self):
if theano.config.device.find('gpu') > -1:
from theano import function, config, shared, sandbox, Out
import theano.tensor as T
import numpy
import time
vlen = 10 * 30 * 70 # 10 x #cores x # threads per core
iters = 10
rng = numpy.random.RandomState(22)
x = shared(numpy.asarray(rng.rand(vlen), config.floatX))
f = function([],
Out(sandbox.cuda.basic_ops.gpu_from_host(T.exp(x)),
borrow=True))
# print f.maker.fgraph.toposort()
t0 = time.time()
for i in xrange(iters):
r = f()
t1 = time.time()
print 'Looping %d times took' % iters, t1 - t0, 'seconds'
print 'Result is', r
print 'Numpy result is', numpy.asarray(r)
if numpy.any([isinstance(x.op, T.Elemwise)
for x in f.maker.fgraph.toposort()]):
print 'Used the cpu'
else:
print 'Used the gpu'
assert not numpy.any([isinstance(x.op, T.Elemwise)
for x in f.maker.fgraph.toposort()])
def test4d(self):
g = Graph()
oper = OpThresholdOneLevel(graph=g)
oper.MinSize.setValue(self.minSize)
oper.MaxSize.setValue(self.maxSize)
oper.Threshold.setValue(0.5)
oper.InputImage.setValue(self.data)
output = oper.Output[:].wait()
assert numpy.all(output.shape == self.data.shape)
clusters = self.generateData((self.nx, self.ny, self.nz))
cluster1 = numpy.logical_and(output, clusters[0])
assert numpy.any(cluster1 != 0)
oper.MinSize.setValue(5)
output = oper.Output[:].wait()
cluster1 = numpy.logical_and(output, clusters[0])
assert numpy.all(cluster1 == 0)
cluster4 = numpy.logical_and(output.squeeze(), clusters[3])
assert numpy.all(cluster4 == 0)
cluster5 = numpy.logical_and(output.squeeze(), clusters[2])
assert numpy.all(cluster5 == 0)
oper.Threshold.setValue(0.2)
output = oper.Output[:].wait()
cluster5 = numpy.logical_and(output.squeeze(), clusters[2])
assert numpy.any(cluster5 != 0)
def pop_planes(geometry, kwargs):
# Convert miller index specifications to normal vectors
miller_defs = kwargs.pop("planes_miller", None)
if miller_defs is not None:
if np.any(np.all(abs(miller_defs[:,0:3]) < EPSILON, axis=1)):
error("Emtpy miller index tuple")
miller_defs[:,0:3] = miller_to_normal(
np.dot(geometry.latvecs, geometry.bravais_cell),
miller_defs[:,0:3])
else:
miller_defs = np.zeros((0, 4), dtype=float)
# Convert plane normal vector specifications into cartesian coords.
normal_defs = kwargs.pop("planes_normal", None)
if normal_defs is not None:
normal_defs[:,0:3] = geometry.coord_transform(
normal_defs[:,0:3],
kwargs.pop("planes_normal_coordsys", "lattice"))
if np.any(np.all(abs(normal_defs[:,0:3]) < EPSILON, axis=1)):
error("Emtpy normal vector definition")
else:
normal_defs = np.zeros((0, 4), dtype=float)
# Append two defintions
planes_normal = np.vstack(( miller_defs, normal_defs ))
return planes_normal
def test_sequential_as_downstream_of_masking_layer(self):
inputs = keras.layers.Input(shape=(3, 4))
x = keras.layers.Masking(mask_value=0., input_shape=(3, 4))(inputs)
s = keras.Sequential()
s.add(keras.layers.Dense(5, input_shape=(4,)))
x = keras.layers.wrappers.TimeDistributed(s)(x)
model = keras.Model(inputs=inputs, outputs=x)
model.compile(
optimizer='rmsprop',
loss='mse',
run_eagerly=testing_utils.should_run_eagerly())
model_input = np.random.randint(
low=1, high=5, size=(10, 3, 4)).astype('float32')
for i in range(4):
model_input[i, i:, :] = 0.
model.fit(model_input,
np.random.random((10, 3, 5)), epochs=1, batch_size=6)
if not context.executing_eagerly():
# Note: this doesn't work in eager due to DeferredTensor/ops compatibility
# issue.
mask_outputs = [model.layers[1].compute_mask(model.layers[1].input)]
mask_outputs += [model.layers[2].compute_mask(
model.layers[2].input, mask_outputs[-1])]
func = keras.backend.function([model.input], mask_outputs)
mask_outputs_val = func([model_input])
self.assertAllClose(mask_outputs_val[0], np.any(model_input, axis=-1))
self.assertAllClose(mask_outputs_val[1], np.any(model_input, axis=-1))
def __call__(self, x, y, xval, bounds_error=True, fill_value=np.nan):
if np.isscalar(xval): # xval is a scalar
if xval < x[0] or xval > x[-1]: # the value is out of bounds
if bounds_error:
raise Exception("x value is out of interpolation bounds")
else:
return np.nan
else: # the value is in the bounds
return self.f(x, y, xval)
else: # xval is an array
inside = (xval >= x[0]) & (xval <= x[-1])
outside = ~inside
if np.any(outside): # some values are out of bounds
if bounds_error:
raise Exception("x values are out of interpolation bounds")
else:
if np.any(inside):
yval = np.zeros(xval.shape)
yval[inside] = self.f(x, y, xval[inside])
yval[outside] = fill_value
return yval
else:
return np.repeat(fill_value, xval.shape)
else: # all values are in the bounds
return self.f(x, y, xval)
def test_impute_random(self):
nan = numpy.nan
data = [
[1.0, nan, 0.0],
[2.0, 1.0, 3.0],
[nan, nan, nan]
]
domain = Orange.data.Domain(
(Orange.data.DiscreteVariable("A", values=["0", "1", "2"]),
Orange.data.ContinuousVariable("B"),
Orange.data.ContinuousVariable("C"))
)
data = Orange.data.Table.from_numpy(domain, numpy.array(data))
cimp1 = column_imputer_random(domain[0], data)
self.assertTrue(not numpy.any(numpy.isnan(cimp1(data).X)))
cimp2 = column_imputer_random(domain[1], data)
self.assertTrue(not numpy.any(numpy.isnan(cimp2(data).X)))
cimp3 = column_imputer_random(domain[2], data)
self.assertTrue(not numpy.any(numpy.isnan(cimp3(data).X)))
imputer = ImputerModel(
data.domain,
{data.domain[0]: cimp1,
data.domain[1]: cimp2,
data.domain[2]: cimp3}
)
idata = imputer(data)
self.assertTrue(not numpy.any(numpy.isnan(idata.X)))
definedmask = ~numpy.isnan(data.X)
self.assertClose(data.X[definedmask],
idata.X[definedmask])
def tracks_to_expected(tracks, vol_dims):
# simulate expected behavior of module
vol_dims = np.array(vol_dims, dtype=np.int32)
counts = np.zeros(vol_dims, dtype=np.int32)
elements = {}
for t_no, t in enumerate(tracks):
u_ps = set()
ti = np.round(t).astype(np.int32)
for p_no, p in enumerate(ti):
if np.any(p < 0):
p[p<0] = 0
too_high = p >= vol_dims
if np.any(too_high):
p[too_high] = vol_dims[too_high]-1
p = tuple(p)
if p in u_ps:
continue
u_ps.add(p)
val = t_no
if counts[p]:
elements[p].append(val)
else:
elements[p] = [val]
counts[p] +=1
return counts, elements
def j_roots(n, alpha, beta, mu=0):
"""[x,w] = j_roots(n,alpha,beta)
Returns the roots (x) of the nth order Jacobi polynomial, P^(alpha,beta)_n(x)
and weights (w) to use in Gaussian Quadrature over [-1,1] with weighting
function (1-x)**alpha (1+x)**beta with alpha,beta > -1.
"""
if any(alpha <= -1) or any(beta <= -1):
raise ValueError("alpha and beta must be greater than -1.")
assert n > 0, "n must be positive."
(p, q) = (alpha, beta)
# from recurrence relations
sbn_J = (
lambda k: 2.0
/ (2.0 * k + p + q)
* sqrt((k + p) * (k + q) / (2 * k + q + p + 1))
* (np.where(k == 1, 1.0, sqrt(k * (k + p + q) / (2.0 * k + p + q - 1))))
)
if any(p == q): # XXX any or all???
an_J = lambda k: 0.0 * k
else:
an_J = lambda k: np.where(
k == 0, (q - p) / (p + q + 2.0), (q * q - p * p) / ((2.0 * k + p + q) * (2.0 * k + p + q + 2))
)
g = cephes.gamma
mu0 = 2.0 ** (p + q + 1) * g(p + 1) * g(q + 1) / (g(p + q + 2))
val = gen_roots_and_weights(n, an_J, sbn_J, mu0)
if mu:
return val + [mu0]
else:
return val
def get_polar_motion(time):
"""
gets the two polar motion components in radians for use with apio13
"""
# Get the polar motion from the IERS table
xp, yp, status = iers.IERS_Auto.open().pm_xy(time, return_status=True)
wmsg = None
if np.any(status == iers.TIME_BEFORE_IERS_RANGE):
wmsg = ('Tried to get polar motions for times before IERS data is '
'valid. Defaulting to polar motion from the 50-yr mean for those. '
'This may affect precision at the 10s of arcsec level')
xp.ravel()[status.ravel() == iers.TIME_BEFORE_IERS_RANGE] = _DEFAULT_PM[0]
yp.ravel()[status.ravel() == iers.TIME_BEFORE_IERS_RANGE] = _DEFAULT_PM[1]
warnings.warn(wmsg, AstropyWarning)
if np.any(status == iers.TIME_BEYOND_IERS_RANGE):
wmsg = ('Tried to get polar motions for times after IERS data is '
'valid. Defaulting to polar motion from the 50-yr mean for those. '
'This may affect precision at the 10s of arcsec level')
xp.ravel()[status.ravel() == iers.TIME_BEYOND_IERS_RANGE] = _DEFAULT_PM[0]
yp.ravel()[status.ravel() == iers.TIME_BEYOND_IERS_RANGE] = _DEFAULT_PM[1]
warnings.warn(wmsg, AstropyWarning)
return xp.to(u.radian).value, yp.to(u.radian).value
def sample_representer_points(self):
# Sample representer points only in the
# configuration space by setting all environmental
# variables to 1
D = np.where(self.is_env == 0)[0].shape[0]
lower = self.lower[np.where(self.is_env == 0)]
upper = self.upper[np.where(self.is_env == 0)]
self.sampling_acquisition.update(self.model)
for i in range(5):
restarts = np.random.uniform(low=lower,
high=upper,
size=(self.Nb, D))
sampler = emcee.EnsembleSampler(self.Nb, D,
self.sampling_acquisition_wrapper)
self.zb, self.lmb, _ = sampler.run_mcmc(restarts, 50)
if not np.any(np.isinf(self.lmb)):
break
else:
print("Infinity")
if np.any(np.isinf(self.lmb)):
raise ValueError("Could not sample valid representer points! LogEI is -infinity")
if len(self.zb.shape) == 1:
self.zb = self.zb[:, None]
if len(self.lmb.shape) == 1:
self.lmb = self.lmb[:, None]
# Project representer points to subspace
proj = np.ones([self.zb.shape[0],
self.upper[self.is_env == 1].shape[0]])
proj *= self.upper[self.is_env == 1].shape[0]
self.zb = np.concatenate((self.zb, proj), axis=1)
def algebraic2parametric(coeff):
'''
Based on matlab function "ellipse_param.m" which accompanies
"Least-Squares Fitting of Circles and Ellipses", W. Gander, G. H. Golub, R. Strebel,
BIT Numerical Mathematics, Springer 1994
convert the coefficients (a,b,c,d,e,f) of the algebraic equation:
ax^2 + bxy + cy^2 + dx + ey + f = 0
to the parameters of the parametric equation. The parameters are
returned as a dictionary containing:
center - center of the ellipse
a - major axis
b - minor axis
alpha - angle of major axis
convention note: alpha is measured as positive values towards the y-axis
'''
#print coeff
#print ("A=%.3f B=%.3f C=%.3f D=%.3f E=%.3f F=%.3f"
# %(coeff[0], coeff[1], coeff[2], coeff[3], coeff[4], coeff[5],))
if numpy.any(numpy.isnan(coeff)) or numpy.any(numpy.isinf(coeff)):
return None
A = numpy.array((coeff[0], coeff[1]/2, coeff[1]/2, coeff[2]))
A.shape = 2,2
bb = numpy.asarray(coeff[3:5])
c = coeff[5]
D,Q = scipy.linalg.eig(A)
D = D.real
det = D[0]*D[1]
if det <= 0:
return None
else:
bs = numpy.dot(Q.transpose(), bb)
alpha = numpy.arctan2(Q[1,0], Q[0,0])
zs = scipy.linalg.solve(-2*numpy.diagflat(D), bs)
z = numpy.dot(Q, zs)
h = numpy.dot(-bs.transpose(), zs) / 2 - c
a = numpy.sqrt(h/D[0])
b = numpy.sqrt(h/D[1])
## correct backwards major/minor axes
## 'major axis as a, minor axis as b'
if b > a:
temp = b
b = a
a = temp
alpha = math.pi/2 + alpha
#print "alpha", alpha
if alpha <= -math.pi/2:
alpha += math.pi
elif alpha > math.pi/2:
alpha -= math.pi
return {'center':z, 'a':a, 'b':b, 'alpha':alpha}
def testBinds(self):
ds = normalFeatureDataset()
ds_data = ds.samples.copy()
ds_chunks = ds.chunks.copy()
self.failUnless(N.all(ds.samples == ds_data)) # sanity check
funcs = ['zscore', 'coarsenChunks']
if externals.exists('scipy'):
funcs.append('detrend')
for f in funcs:
eval('ds.%s()' % f)
self.failUnless(N.any(ds.samples != ds_data) or
N.any(ds.chunks != ds_chunks),
msg="We should have modified original dataset with %s" % f)
ds.samples = ds_data.copy()
ds.chunks = ds_chunks.copy()
# and some which should just return results
for f in ['aggregateFeatures', 'removeInvariantFeatures',
'getSamplesPerChunkLabel']:
res = eval('ds.%s()' % f)
self.failUnless(res is not None,
msg='We should have got result from function %s' % f)
self.failUnless(N.all(ds.samples == ds_data),
msg="Function %s should have not modified original dataset" % f)
def collide(self):
for a in self.actors:
a.collision_prepare()
for i, ai in enumerate(self.actors):
if isinstance(ai, ParticleActor):
ai.collision_self()
for j, aj in enumerate(self.actors):
if isinstance(aj, ParticleActor):
continue
if i == j: continue
info = spatial.Info(ai.spatial_grid, aj.spatial_mesh, i==j)
mask = info.triangle != -1
active = np.flatnonzero(mask) # active vertex idx
if np.any(mask):# and not isinstance(ai, StaticActor):
triangle = info.triangle[active]
bary = info.bary[active]
velocity = aj.velocity[aj.mesh.faces[triangle]]
velocity = np.einsum('vtc,vt->vc', velocity, bary)
relative_velocity = ai.velocity[active] - velocity
friction = 1e-2
stiffness = 1e1
force = info.depth[active][:, None] * info.normal[active] * stiffness - relative_velocity * friction
assert not np.any(np.isnan(force))
np.add.at(ai.force, active, force)
corners = aj.mesh.faces[triangle]
for i in range(3):
np.add.at(aj.force, corners[:, i], -bary[:, [i]] * force)
def backward(self, top, propagate_down, bottom):
h=bottom[0].data.shape[2]
w=bottom[0].data.shape[3]
num_proposals = bottom[1].data.shape[0]
fbox = self.fbox
bottom[0].diff[...] = 0
for b in np.arange(num_proposals):
if np.any(self.act[b]==True):
wb=fbox[b,2]-fbox[b,0]
hb=fbox[b,3]-fbox[b,1]
if hb==0 or wb==0: #bounding box with size 0
pass
elif h==hb and w==wb: #bounding box with size of the image
pass
else:
bottom[0].diff[0,self.act[b],:,:] += bottom[1].data[b,self.act[b],np.newaxis,np.newaxis]/(h*w-hb*wb) #negative
bottom[0].diff[0,self.act[b],fbox[b,1]:fbox[b,3]+1,fbox[b,0]:fbox[b,2]+1] += -bottom[1].data[b,self.act[b],np.newaxis,np.newaxis]/(h*w-hb*wb) #negative
bottom[0].diff[0,self.act[b],fbox[b,1]:fbox[b,3]+1,fbox[b,0]:fbox[b,2]+1] += -bottom[1].data[b,self.act[b],np.newaxis,np.newaxis]/(hb*wb) #positive
bottom[0].diff[...] *= self.myloss_weight
bottom[1].diff[...] = self.myloss_weight*self.hinge
if np.any(np.isnan(bottom[0].diff)):
print "Nan error"
dsfsf
if np.any(np.isnan(bottom[1].diff)):
print "Nan error"
dsfsf
def check_obs_scheme(self):
" Checks the internal validity of provided observation schemes "
# check sub_pops
idx_union = np.sort(self._sub_pops[0])
i = 1
while idx_union.size < self._p and i < len(self._sub_pops):
idx_union = np.union1d(idx_union, self._sub_pops[i])
i += 1
if idx_union.size != self._p or np.any(idx_union!=np.arange(self._p)):
raise Exception(('all subpopulations together have to cover '
'exactly all included observed varibles y_i in y.'
'This is not the case. Change the difinition of '
'subpopulations in variable sub_pops or reduce '
'the number of observed variables p. '
'The union of indices of all subpopulations is'),
idx_union )
# check obs_time
if not self._obs_time[-1]==self._T:
raise Exception(('Entries of obs_time give the respective ends of '
'the periods of observation for any '
'subpopulation. Hence the last entry of obs_time '
'has to be the full recording length. The last '
'entry of obs_time before is '), self._obs_time[-1])
if np.any(np.diff(self._obs_time)<1):
raise Exception(('lengths of observation have to be at least 1. '
'Minimal observation time for a subpopulation: '),
np.min(np.diff(self._obs_time)))
# check obs_pops
if not self._obs_time.size == self._obs_pops.size:
raise Exception(('each entry of obs_pops gives the index of the '
'subpopulation observed up to the respective '
'time given in obs_time. Thus the sizes of the '
'two arrays have to match. They do not. '
'no. of subpop. switch points and no. of '
'subpopulations ovserved up to switch points '
'are '), (self._obs_time.size, self._obs_pops.size))
idx_pops = np.sort(np.unique(self._obs_pops))
if not np.min(idx_pops)==0:
raise Exception(('first subpopulation has to have index 0, but '
'is given the index '), np.min(idx_pops))
elif not idx_pops.size == len(self._sub_pops):
raise Exception(('number of specified subpopulations in variable '
'sub_pops does not meet the number of '
'subpopulations indexed in variable obs_pops. '
'Delete subpopulations that are never observed, '
'or change the observed subpopulations in '
'variable obs_pops accordingly. The number of '
'indexed subpopulations is '),
len(self._sub_pops))
elif not np.all(np.diff(idx_pops)==1):
raise Exception(('subpopulation indices have to be consecutive '
'integers from 0 to the total number of '
'subpopulations. This is not the case. '
'Given subpopulation indices are '),
idx_pops)
def test_unwrap(self):
"""Test different geometry types are appropriately unwrapped."""
wrapper = Wrapper()
path = tempfile.mkdtemp()
for desc, geom in self.possible.iteritems():
unwrapped = wrapper.unwrap(geom)
if desc in self.actual_unwrapped:
self.assertTrue(self.actual_unwrapped[desc].almost_equals(unwrapped, decimal=5))
try:
self.assertEqual(type(geom), type(unwrapped))
except AssertionError:
if desc == 'axis_polygon':
# by necessity of being split on the axis, this will come out as a multipolygon
self.assertIsInstance(unwrapped, MultiPolygon)
else:
raise
self.assertFalse(np.any(np.array(unwrapped) < 0.0))
if isinstance(unwrapped, (MultiPolygon, Polygon)):
it = get_iter(unwrapped)
for polygon in it:
self.assertFalse(np.any(np.array(polygon.exterior) > 360.0))
else:
self.assertFalse(np.any(np.array(unwrapped) > 360.0))
def process_shoreline(contours, cloud_mask, georef, image_epsg, settings):
"""
Converts the contours from image coordinates to world coordinates.
This function also removes the contours that are too small to be a shoreline
(based on the parameter settings['min_length_sl'])
KV WRL 2018
Arguments:
-----------
contours: np.array or list of np.array
image contours as detected by the function find_contours
cloud_mask: np.array
2D cloud mask with True where cloud pixels are
georef: np.array
vector of 6 elements [Xtr, Xscale, Xshear, Ytr, Yshear, Yscale]
image_epsg: int
spatial reference system of the image from which the contours were extracted
settings: dict with the following keys
'output_epsg': int
output spatial reference system
'min_length_sl': float
minimum length of shoreline contour to be kept (in meters)
Returns:
-----------
shoreline: np.array
array of points with the X and Y coordinates of the shoreline
"""
# convert pixel coordinates to world coordinates
contours_world = SDS_tools.convert_pix2world(contours, georef)
# convert world coordinates to desired spatial reference system
contours_epsg = SDS_tools.convert_epsg(contours_world, image_epsg, settings['output_epsg'])
# remove contours that have a perimeter < min_length_sl (provided in settings dict)
# this enables to remove the very small contours that do not correspond to the shoreline
contours_long = []
for l, wl in enumerate(contours_epsg):
coords = [(wl[k,0], wl[k,1]) for k in range(len(wl))]
a = LineString(coords) # shapely LineString structure
if a.length >= settings['min_length_sl']:
contours_long.append(wl)
# format points into np.array
x_points = np.array([])
y_points = np.array([])
for k in range(len(contours_long)):
x_points = np.append(x_points,contours_long[k][:,0])
y_points = np.append(y_points,contours_long[k][:,1])
contours_array = np.transpose(np.array([x_points,y_points]))
shoreline = contours_array
# now remove any shoreline points that are attached to cloud pixels
if sum(sum(cloud_mask)) > 0:
# get the coordinates of the cloud pixels
idx_cloud = np.where(cloud_mask)
idx_cloud = np.array([(idx_cloud[0][k], idx_cloud[1][k]) for k in range(len(idx_cloud[0]))])
# convert to world coordinates and same epsg as the shoreline points
coords_cloud = SDS_tools.convert_epsg(SDS_tools.convert_pix2world(idx_cloud, georef),
image_epsg, settings['output_epsg'])[:,:-1]
# only keep the shoreline points that are at least 30m from any cloud pixel
idx_keep = np.ones(len(shoreline)).astype(bool)
for k in range(len(shoreline)):
if np.any(np.linalg.norm(shoreline[k,:] - coords_cloud, axis=1) < 30):
idx_keep[k] = False
shoreline = shoreline[idx_keep]
return shoreline
def __init__(self,
loss_factory,
X,
penalty_structure=None,
group_weights={},
elastic_net=iq(0, 0, 0, 0),
alpha=0.,
intercept=True,
positive_part=None,
unpenalized=None,
lagrange_proportion=0.05,
nstep=100,
scale=True,
center=True):
self.loss_factory = loss_factory
self.scale = scale
self.center = center
# for group lasso weights, if implied by penalty_structure
self.group_weights = group_weights
# normalize X, adding intercept if needed
self.intercept = intercept
p = X.shape[1]
if self.intercept:
self.penalty_structure = np.ones(p + 1) * L1_PENALTY
self.penalty_structure[0] = UNPENALIZED
if penalty_structure is not None:
self.penalty_structure[1:] = penalty_structure
if scipy.sparse.issparse(X):
self._X1 = scipy.sparse.hstack([np.ones((X.shape[0], 1)),
X]).tocsc()
else:
self._X1 = np.hstack([np.ones((X.shape[0], 1)), X])
if self.scale or self.center:
self._Xn = normalize(self._X1,
center=self.center,
scale=self.scale,
intercept_column=0)
which_0 = self._Xn.col_stds == 0
else:
self._Xn = self._X1
which_0 = np.zeros(self._Xn.shape)
else:
self.penalty_structure = np.ones(p) * L1_PENALTY
if penalty_structure is not None:
self.penalty_structure[:] = penalty_structure
if self.scale or self.center:
self._Xn = normalize(X, center=self.center, scale=self.scale)
which_0 = self._Xn.col_stds == 0
else:
self._Xn = X
which_0 = np.zeros(self._Xn.shape)
if np.any(which_0):
self._selector = selector(~which_0, self._Xn.input_shape)
if self.scale or self.center:
self._Xn = self._Xn.slice_columns(~which_0)
else:
self._Xn = self._Xn[:, ~which_0]
else:
if self.scale or self.center:
self._selector = identity(self._Xn.input_shape)
else:
self._selector = identity(self._Xn.shape)
# the penalty parameters
self.alpha = alpha
self.lagrange_proportion = lagrange_proportion
self.nstep = nstep
self._elastic_net = elastic_net.collapsed()
self.initial_active = (np.equal(self.penalty_structure, UNPENALIZED) +
np.equal(self.penalty_structure, NONNEGATIVE))
self.ever_active = self.initial_active.copy()
def main(self, inner_tol=1.e-5, verbose=False):
# scaling will be needed to get coefficients on original scale
if self.scale:
scalings = np.asarray(self.Xn.col_stds).reshape(-1)
else:
scalings = np.ones(self.shape[1])
scalings = self.nonzero.adjoint_map(scalings)
# take a guess at the inverse step size
self.final_step = 1000. / self.lipschitz
lseq = self.lagrange_sequence # shorthand
# first solution corresponding to all zeros except intercept
self.solution[:] = self.null_solution.copy()
grad_solution = self.grad().copy()
strong, strong_selector = self.strong_set(lseq[0],
lseq[1],
grad=grad_solution)
p = self.shape[0]
rescaled_solutions = scipy.sparse.csr_matrix(
self.nonzero.adjoint_map(self.solution) / scalings)
objective = [self.loss.smooth_objective(self.solution, 'func')]
# not quite right -- should check tight constraints
dfs = [np.sum(self.initial_active)]
retry_counter = 0
all_failing = np.zeros(grad_solution.shape, np.bool)
for lagrange_new, lagrange_cur in zip(lseq[1:], lseq[:-1]):
self.lagrange = lagrange_new
tol = inner_tol
active_old = self.active.copy()
num_tries = 0
debug = False
coef_stop = True
while True:
strong, strong_selector = self.strong_set(lagrange_cur,
lagrange_new,
grad=grad_solution)
subproblem_set = self.ever_active + all_failing
final_step, grad, sub_soln, penalty_structure \
= self.solve_subproblem(subproblem_set,
lagrange_new,
tol=tol,
start_step=self.final_step,
debug=debug and verbose,
coef_stop=coef_stop)
p = self.shape[1]
self.solution[subproblem_set][:] = sub_soln
# this only corrects the gradient on the subproblem_set
grad_solution[subproblem_set][:] = grad
strong_problem = self.restricted_problem(strong,
lagrange_new)[0]
strong_soln = self.solution[strong]
strong_grad = (
strong_problem.smooth_objective(strong_soln, mode='grad') +
self.elastic_net[strong].objective(strong_soln,
mode='grad'))
strong_penalty = strong_problem.proximal_atom
strong_failing = check_KKT(strong_penalty, strong_grad,
strong_soln, lagrange_new)
if np.any(strong_failing):
all_failing += (strong_selector.adjoint_map(strong_failing)
!= 0)
else:
self.solution[subproblem_set][:] = sub_soln
grad_solution = self.grad()
all_failing = check_KKT(self.penalty, grad_solution,
self.solution, lagrange_new)
if not all_failing.sum():
self.ever_active += self.solution != 0
self.final_step = final_step
break
else:
if verbose:
print('failing:', np.nonzero(all_failing)[0])
retry_counter += 1
self.ever_active += all_failing
tol /= 2.
num_tries += 1
if num_tries % 5 == 0:
self.solution[subproblem_set][:] = sub_soln
self.solution[~subproblem_set][:] = 0
grad_solution = self.grad()
debug = True
tol = inner_tol
if num_tries >= 10:
warn('convergence not achieved for lagrange=%0.4e' %
lagrange_new)
break
rescaled_solution = self.nonzero.adjoint_map(self.solution)
rescaled_solutions = scipy.sparse.vstack(
[rescaled_solutions, rescaled_solution])
objective.append(
self.loss.smooth_objective(self.solution, mode='func'))
dfs.append(self.ever_active.shape[0])
gc.collect()
if verbose:
print(lagrange_cur / self.lagrange_max, lagrange_new,
(self.solution != 0).sum(),
1. - objective[-1] / objective[0],
list(self.lagrange_sequence).index(lagrange_new),
np.fabs(rescaled_solution).sum())
objective = np.array(objective)
output = {
'devratio': 1 - objective / objective.max(),
'df': dfs,
'lagrange': self.lagrange_sequence,
'scalings': scalings,
'beta': rescaled_solutions.T
}
return output
change_Lr = tf.keras.callbacks.LearningRateScheduler(scheduler)
print("OPTIMIZER\n")
#select optimizer，选择优化器
optimizer = tf.keras.optimizers.Adam(learning_rate=learning_rate,
decay=decay_rate)
#checkpoint_callback = tf.keras.callbacks.ModelCheckpoint(
# weight_path, monitor='val_accuracy', verbose=1,
# save_best_only=False, save_weights_only=True,
# save_frequency=1 )
Model.compile(optimizer=optimizer,
loss=tf.keras.losses.categorical_crossentropy,
metrics=['accuracy'])
print("Model fit\n")
print('is there any NaN in x_train? ', np.any(np.isnan(x_train)))
print('is there any NaN in y_train? ', np.any(np.isnan(y_train)))
training = Model.fit(x_train,
one_hot_y_train,
batch_size=BATCH_SIZE,
validation_data=[x_validation, one_hot_y_validation],
epochs=EPOCHS,
callbacks=[change_Lr]) #
print(training.history)
#To visualize the training process
def plot_history(training):
plt.plot(training.history['loss'])
plt.plot(training.history['val_loss'])
def actor_critic(agent_name,
multiple_agents=False,
load_agent=False,
n_episodes=300,
max_t=1000,
train_mode=True):
""" Batch processed the states in a single forward pass with a single neural network
Params
======
multiple_agents (boolean): boolean for multiple agents
PER (boolean):
n_episodes (int): maximum number of training episodes
max_t (int): maximum number of timesteps per episode
"""
start = time.time()
device = get_device()
env, env_info, states, state_size, action_size, brain_name, num_agents = initialize_env(
multiple_agents, train_mode)
states = torch.from_numpy(states).to(device).float()
NUM_PROCESSES = num_agents
# Scores is Episode Rewards
scores = np.zeros(num_agents)
scores_window = deque(maxlen=100)
scores_episode = []
actor_critic = ActorCritic(state_size, action_size, device).to(device)
agent = A2C_ACKTR(agent_name,
actor_critic,
value_loss_coef=CRITIC_DISCOUNT,
entropy_coef=ENTROPY_BETA,
lr=LEARNING_RATE,
eps=EPS,
alpha=ALPHA,
max_grad_norm=MAX_GRAD_NORM,
acktr=False,
load_agent=load_agent)
rollouts = SimpleRolloutStorage(NUM_STEPS, NUM_PROCESSES, state_size,
action_size)
rollouts.to(device)
num_updates = NUM_ENV_STEPS // NUM_STEPS // NUM_PROCESSES
# num_updates = NUM_ENV_STEPS // NUM_STEPS
print("\n## Loaded environment and agent in {} seconds ##\n".format(
round((time.time() - start), 2)))
update_start = time.time()
timesteps = 0
episode = 0
if load_agent != False:
episode = agent.episode
while True:
"""CAN INSERT LR DECAY HERE"""
# if episode == MAX_EPISODES:
# return scores_episode
# Adds noise to agents parameters to encourage exploration
# agent.add_noise(PARAMETER_NOISE)
for step in range(NUM_STEPS):
step_start = time.time()
# Sample actions
with torch.no_grad():
values, actions, action_log_probs, _ = agent.act(states)
clipped_actions = np.clip(actions.cpu().numpy(), *ACTION_BOUNDS)
env_info = env.step(actions.cpu().numpy())[
brain_name] # send the action to the environment
next_states = env_info.vector_observations # get the next state
rewards = env_info.rewards # get the reward
rewards_tensor = np.array(env_info.rewards)
rewards_tensor[rewards_tensor == 0] = NEGATIVE_REWARD
rewards_tensor = torch.from_numpy(rewards_tensor).to(
device).float().unsqueeze(1)
dones = env_info.local_done
masks = torch.from_numpy(1 - np.array(dones).astype(int)).to(
device).float().unsqueeze(1)
rollouts.insert(states, actions, action_log_probs, values,
rewards_tensor, masks, masks)
next_states = torch.from_numpy(next_states).to(device).float()
states = next_states
scores += rewards
# print(rewards)
if timesteps % 100:
print('\rTimestep {}\tScore: {:.2f}\tmin: {:.2f}\tmax: {:.2f}'.
format(timesteps, np.mean(scores), np.min(scores),
np.max(scores)),
end="")
if np.any(dones):
print(
'\rEpisode {}\tScore: {:.2f}\tAverage Score: {:.2f}\tMin Score: {:.2f}\tMax Score: {:.2f}'
.format(episode, score, np.mean(scores_window),
np.min(scores), np.max(scores)),
end="\n")
update_csv(agent_name, episode, np.mean(scores_window),
np.max(scores))
if episode % 20 == 0:
agent.save_agent(agent_name,
score,
episode,
save_history=True)
else:
agent.save_agent(agent_name, score, episode)
episode += 1
scores = np.zeros(num_agents)
break
timesteps += 1
with torch.no_grad():
next_values, _, _, _ = agent.act(next_states)
rollouts.compute_returns(next_values, USE_GAE, GAMMA, GAE_LAMBDA)
agent.update(rollouts)
score = np.mean(scores)
scores_window.append(score) # save most recent score
scores_episode.append(score)
return scores_episode
def stepp_analysis(year, mw, dm=0.1, dt=1, ttol=0.2, iloc=True):
"""
Stepp algorithm
:param year: catalog matrix year column
:type year: numpy.ndarray
:param mw: catalog matrix magnitude column
:type mw: numpy.ndarray
:keyword dm: magnitude interval/window
:type dm: positive float
:keyword dt: time interval
:type dt: int
:keyword ttol: tolerance threshold
:type ttol: positive float
:keyword iloc: Fix analysis such that completeness magnitude
can only increase with catalogue duration
(i.e. completess cannot increase for more recent
catalogues)
:type iloc: bool
:returns: two-column completeness table representing the earliest
year at which the catalogue is complete above a
given magnitude
:rtype: numpy.ndarray
"""
# Round off the magnitudes to 2 d.p
mw = np.around(100.0 * mw) / 100.0
lowm = np.floor(10. * np.min(mw)) / 10.
highm = np.ceil(10. * np.max(mw)) / 10.
# Determine magnitude bins
mbin = np.arange(lowm, highm + dm, dm)
ntb = np.max(np.shape(mbin))
# Determine time bins
end_time = np.max(year)
start_time = np.min(year)
time_range = np.arange(dt, end_time - start_time + 2, dt)
nt = np.max(np.shape(time_range))
t_upper_bound = end_time * np.ones(nt)
t_lower_bound = t_upper_bound - time_range
t_rate = 1. / np.sqrt(time_range) # Poisson rate
number_obs = np.zeros((nt, ntb - 1))
lamda = np.zeros((nt, ntb - 1))
siglam = np.zeros((nt, ntb - 1))
ii = 0
# count number of events catalogue and magnitude windows
while ii <= (nt - 1):
# Select earthquakes later than or in Year[ii]
yrchk = year >= t_lower_bound[ii]
mtmp = mw[yrchk]
jj = 0
while jj <= (ntb - 2):
#Count earthquakes in magnitude bin
if jj == (ntb - 2):
number_obs[ii, jj] = np.sum(mtmp >= np.sum(mbin[jj]))
else:
number_obs[ii, jj] = np.sum(
np.logical_and(mtmp >= mbin[jj], mtmp < mbin[jj + 1]))
jj = jj + 1
ii = ii + 1
time_diff = (np.log10(t_rate[1:]) - np.log10(t_rate[:-1]))
time_diff = time_diff / (np.log10(time_range[1:]) -
np.log10(time_range[:-1]))
comp_length = np.zeros((ntb - 1, 1))
tloc = np.zeros((ntb - 1, 1), dtype=int)
ii = 0
while ii < (ntb - 1):
lamda[:, ii] = number_obs[:, ii] / time_range
siglam[:, ii] = np.sqrt(lamda[:, ii] / time_range)
zero_find = siglam[:, ii] < 1E-14 # To avoid divide by zero
siglam[zero_find, ii] = 1E-14
grad1 = (np.log10(siglam[1:, ii]) - np.log10(siglam[:-1, ii]))
grad1 = grad1 / (np.log10(time_range[1:]) - np.log10(time_range[:-1]))
resid1 = grad1 - time_diff
test1 = np.abs(resid1[1:] - resid1[:-1])
tloct = np.nonzero(test1 > ttol)[0]
if not (np.any(tloct)):
tloct = -1
else:
tloct = tloct[-1]
if tloct < 0:
# No location passes test
if ii > 0:
# Use previous value
tloc[ii] = tloc[ii - 1]
else:
# Print warning
LOGGER.critical(
"Fitting tolerance removed all data - change parameter")
else:
tloc[ii] = tloct
if tloct > np.max(np.shape(time_range)):
tloc[ii] = np.max(np.shape(time_range))
if ii > 0:
# If the increasing completeness is option is set
# and the completeness is lower than the previous value
# then fix at previous value
inc_check = np.logical_and(iloc, (tloc[ii] < tloc[ii - 1]))
if inc_check:
tloc[ii] = tloc[ii - 1]
comp_length[ii] = time_range[tloc[ii]]
ii = ii + 1
completeness_table = np.column_stack([end_time - comp_length, mbin[:-1].T])
return completeness_table
def core_test_convolution_double_backward(inshape, kernel, outmaps, pad, stride,
dilation, group, channel_last, with_bias, base_axis, seed, ctx,
func_name, non_accum_check=True,
atol_f=1e-4, atol_b=1e-3, atol_accum=8e-2, dstep=1e-3):
from nbla_test_utils import backward_function_tester, grad_function_forward_function_output
from nnabla.backward_function.convolution import ConvolutionDataGrad, ConvolutionFilterGrad
if func_name == 'ConvolutionCuda':
pytest.skip('CUDA Convolution N-D is only supported in CUDNN extension')
if channel_last and not func_name.endswith('Cudnn'):
pytest.skip(
'channel_last=True is only supported in CUDNN backend so far.')
if channel_last and func_name.endswith('Cudnn') and (np.any(np.asarray(dilation) > 1) or group > 1):
import nnabla_ext.cuda as nc
major, minor, revision = map(int, nc.__cudnn_version__.split('.'))
version = major * 1000 + minor * 100
if version < 7200:
pytest.skip(
'channel_last dilated convolution not work in CUDNN {}.'.format(version))
# base_axis = len(inshape) - len(kernel) - 1
inmaps = inshape[base_axis]
if channel_last:
t = refs.ChannelLastToFirstTranspose(len(inshape), len(kernel))
inshape = tuple(inshape[i] for i in t.inv_axes)
rng = np.random.RandomState(seed)
i = np.clip(rng.randn(*inshape).astype(np.float32), -0.8, 0.8)
kshape = (outmaps,) + (inmaps // group,) + kernel
if channel_last:
t = refs.ChannelLastToFirstTranspose(len(kshape), len(kernel))
kshape = tuple(kshape[i] for i in t.inv_axes)
k = np.clip(rng.randn(*kshape).astype(np.float32), -0.8, 0.8)
b = None
if with_bias:
b = np.clip(rng.randn(outmaps).astype(np.float32), -0.8, 0.8)
inputs = [i, k, b]
atol_half = 1.0 if inmaps > 64 else 1e-1
func_args = [base_axis, pad, stride, dilation, group, channel_last]
# Convolution
backward_function_tester(rng, F.convolution, inputs,
func_args=func_args,
atol_f=atol_f, atol_accum=atol_accum, dstep=dstep,
ctx=ctx)
# DataGrad
df, y = grad_function_forward_function_output(ConvolutionDataGrad,
F.convolution,
ctx, inputs, *func_args)
df.xshape = i.shape
ginputs = [rng.randn(*y.shape), k]
backward_function_tester(rng, df, ginputs,
func_args=[],
atol_f=atol_f, atol_b=atol_b, atol_accum=atol_accum, dstep=dstep,
ctx=ctx, non_accum_check=non_accum_check)
# FilterGrad
df, y = grad_function_forward_function_output(ConvolutionFilterGrad,
F.convolution,
ctx, inputs, *func_args)
df.wshape = k.shape
ginputs = [rng.randn(*y.shape), i]
backward_function_tester(rng, df, ginputs,
func_args=[],
atol_f=atol_f, atol_b=atol_b, atol_accum=atol_accum, dstep=dstep,
ctx=ctx, non_accum_check=non_accum_check)
def classify_image_NN(im_ms, im_extra, cloud_mask, min_beach_area, clf):
"""
Classifies every pixel in the image in one of 4 classes:
- sand --> label = 1
- whitewater (breaking waves and swash) --> label = 2
- water --> label = 3
- other (vegetation, buildings, rocks...) --> label = 0
The classifier is a Neural Network that is already trained.
KV WRL 2018
Arguments:
-----------
im_ms: np.array
Pansharpened RGB + downsampled NIR and SWIR
im_extra:
only used for Landsat 7 and 8 where im_extra is the panchromatic band
cloud_mask: np.array
2D cloud mask with True where cloud pixels are
min_beach_area: int
minimum number of pixels that have to be connected to belong to the SAND class
clf: joblib object
pre-trained classifier
Returns:
-----------
im_classif: np.array
2D image containing labels
im_labels: np.array of booleans
3D image containing a boolean image for each class (im_classif == label)
"""
# calculate features
vec_features = calculate_features(im_ms, cloud_mask, np.ones(cloud_mask.shape).astype(bool))
vec_features[np.isnan(vec_features)] = 1e-9 # NaN values are create when std is too close to 0
# remove NaNs and cloudy pixels
vec_cloud = cloud_mask.reshape(cloud_mask.shape[0]*cloud_mask.shape[1])
vec_nan = np.any(np.isnan(vec_features), axis=1)
vec_mask = np.logical_or(vec_cloud, vec_nan)
vec_features = vec_features[~vec_mask, :]
# classify pixels
labels = clf.predict(vec_features)
# recompose image
vec_classif = np.nan*np.ones((cloud_mask.shape[0]*cloud_mask.shape[1]))
vec_classif[~vec_mask] = labels
im_classif = vec_classif.reshape((cloud_mask.shape[0], cloud_mask.shape[1]))
# create a stack of boolean images for each label
im_sand = im_classif == 1
im_swash = im_classif == 2
im_water = im_classif == 3
# remove small patches of sand or water that could be around the image (usually noise)
im_sand = morphology.remove_small_objects(im_sand, min_size=min_beach_area, connectivity=2)
im_water = morphology.remove_small_objects(im_water, min_size=min_beach_area, connectivity=2)
im_labels = np.stack((im_sand,im_swash,im_water), axis=-1)
return im_classif, im_labels
def run(model, experiment_args, train=True):
total_time_start = time.time()
env, comm_env, memory, ounoise, comm_ounoise, config, summaries, saver, start_episode = experiment_args
start_episode = start_episode if train else 0
NUM_EPISODES = config['n_episodes'] if train else config['n_episodes_test']
EPISODE_LENGTH = config['episode_length'] if train else config[
'episode_length_test']
t = 0
episode_rewards_all = []
episode_aux_rewards_all = []
episode_comm_all = []
episode_comm_any_all = []
print('oops changed')
for i_episode in range(start_episode, NUM_EPISODES):
episode_time_start = time.time()
frames = []
episode_comm = 0
episode_comm_any = 0
hier_C = config['hierarchical_time_scale'] if train else 1
# Initialize the environment and state
observations = np.stack(env.reset())
observations = K.tensor(observations, dtype=K.float32).unsqueeze(1)
cumm_rewards = K.zeros((model.num_agents, 1, 1), dtype=dtype)
ounoise.scale = get_noise_scale(i_episode, config)
comm_ounoise.scale = get_noise_scale(i_episode, config)
# monitoring variables
episode_rewards = np.zeros((model.num_agents, 1, 1))
episode_aux_rewards = np.zeros((model.num_agents, 1, 1))
for i_step in range(EPISODE_LENGTH):
model.to_cpu()
if model.communication == 'hierarchical':
if i_step % hier_C == 0:
observations_init = observations.clone()
cumm_rewards = K.zeros((model.num_agents, 1, 1),
dtype=dtype)
comm_actions = []
for i in range(model.num_agents):
if model.discrete_comm:
comm_action = model.select_comm_action(
observations_init[[i], ], i,
True if train else False)
else:
comm_action = model.select_comm_action(
observations_init[[i], ], i,
comm_ounoise if train else False)
comm_actions.append(comm_action)
comm_actions = K.stack(comm_actions)
#medium = observations_init[(comm_actions > .5)[:,0,0]]
#medium = (K.mean(observations_init, dim=0) if medium.shape == K.Size([0]) else K.mean(medium, dim=0)).unsqueeze(0)
if config['agent_alg'] == 'MAHCDDPG_Multi':
medium = []
allocaters = []
for i in range(model.num_agents):
medium.append(observations_init[to_onehot(
comm_actions[:, :, i])].squeeze(0))
allocaters.append(
np.asarray(comm_actions[:, :,
i].view(-1).argmax()))
allocaters = np.asarray(allocaters)
medium = K.stack(medium)
else:
if model.discrete_comm:
medium = observations_init[to_onehot(
comm_actions[:, :, 1].unsqueeze(2))]
else:
medium = observations_init[to_onehot(comm_actions)]
allocaters = comm_actions.view(-1).argmax()
allocaters = np.asarray(allocaters)
else:
if config['agent_alg'] == 'MAMDDPG':
medium = model.select_comm_action(observations).unsqueeze(
0)
#if i_episode > 1000:
#pdb.set_trace()
actions = []
for i in range(model.num_agents):
if config['agent_alg'] == 'MAHCDDPG_Multi':
action = model.select_action(
K.cat([observations[[i], ], medium[[i], ]], dim=-1), i,
ounoise if train else False)
else:
action = model.select_action(
K.cat([observations[[i], ], medium], dim=-1), i,
ounoise if train else False)
actions.append(action)
actions = K.stack(actions)
next_observations, rewards, dones, infos = env.step(
actions.squeeze(1))
next_observations = K.tensor(next_observations,
dtype=dtype).unsqueeze(1)
rewards = K.tensor(rewards, dtype=dtype).view(-1, 1, 1)
# different types od aux reward to train the second policy
if config['agent_alg'] == 'MAHCDDPG_Multi':
intr_rewards = intrinsic_reward_multi(env, medium.numpy())
else:
intr_rewards = intrinsic_reward(env, medium.numpy())
intr_rewards = K.tensor(intr_rewards, dtype=dtype).view(-1, 1, 1)
cumm_rewards += rewards
if config['aux_reward_type'] == 'intrinsic':
aux_rewards = intr_rewards
elif config['aux_reward_type'] == 'cummulative':
aux_rewards = rewards
# for monitoring
episode_rewards += rewards
episode_aux_rewards += aux_rewards
# if it is the last step we don't need next obs
if i_step == EPISODE_LENGTH - 1:
next_observations = None
# Store the transition in memory
if train:
memory[0].push(observations, actions, next_observations,
aux_rewards, medium, None, None, None, None)
if model.communication == 'hierarchical' and (i_step +
1) % hier_C == 0:
memory[1].push(observations_init, None, next_observations,
cumm_rewards, medium, comm_actions, None,
None, None)
# Move to the next state
observations = next_observations
t += 1
# Use experience replay and train the model
critic_losses = None
actor_losses = None
medium_loss = None
if train:
if (sum([
True for i_memory in memory
if len(i_memory) > config['batch_size'] - 1
]) == len(memory) and t % config['steps_per_update'] == 0):
model.to_cuda()
critic_losses = []
actor_losses = []
for i in range(env.n):
batch = Transition_Comm(*zip(
*memory[0].sample(config['batch_size'])))
if model.communication == 'hierarchical':
batch2 = Transition_Comm(*zip(
*memory[1].sample(config['batch_size'])))
critic_loss, actor_loss = model.update_parameters(
batch, batch2, i)
else:
critic_loss, actor_loss, medium_loss = model.update_parameters(
batch, i)
critic_losses.append(critic_loss)
actor_losses.append(actor_loss)
# Record frames
if config['render'] > 0 and i_episode % config['render'] == 0:
if config['env_id'] == 'waterworld':
frames.append(sc.misc.imresize(env.render(), (300, 300)))
else:
frames.append(env.render(mode='rgb_array')[0])
if config['render_color_change']:
for i_geoms, geoms in enumerate(env.render_geoms):
if i_geoms == env.world.leader:
geoms.set_color(0.85, 0.35, 0.35, 0.55)
else:
geoms.set_color(0.35, 0.35, 0.85, 0.55)
if i_geoms == env.n - 1:
break
if config['agent_alg'] == 'MAHCDDPG_Multi':
episode_comm += np.all((env.world.leader - 1) %
model.num_agents == allocaters)
episode_comm_any += np.any((env.world.leader - 1) %
model.num_agents == allocaters)
elif config['agent_alg'] == 'MAHCDDPG':
episode_comm += np.all(env.world.leader == allocaters)
episode_comm_any += np.any(env.world.leader == allocaters)
# <-- end loop: i_step
### MONITORIRNG ###
episode_rewards_all.append(episode_rewards.sum())
episode_aux_rewards_all.append(episode_aux_rewards.sum())
episode_comm_all.append(episode_comm)
episode_comm_any_all.append(episode_comm_any)
if config['verbose'] > 0:
# Printing out
if (i_episode + 1) % 100 == 0:
print("==> Episode {} of {}".format(i_episode + 1,
NUM_EPISODES))
print(' | Id exp: {}'.format(config['exp_id']))
print(' | Exp description: {}'.format(config['exp_descr']))
print(' | Env: {}'.format(config['env_id']))
print(' | Process pid: {}'.format(config['process_pid']))
print(' | Tensorboard port: {}'.format(config['port']))
print(' | Episode total reward: {}'.format(
episode_rewards.sum()))
print(' | Running mean of total reward: {}'.format(
running_mean(episode_rewards_all)[-1]))
print(' | Running mean of total comm_reward: {}'.format(
running_mean(episode_aux_rewards_all)[-1]))
print(' | Medium loss: {}'.format(medium_loss))
print(' | Time episode: {}'.format(time.time() -
episode_time_start))
print(' | Time total: {}'.format(time.time() -
total_time_start))
if config['verbose'] > 0:
ep_save = i_episode + 1 if (i_episode == NUM_EPISODES -
1) else None
is_best_save = None
is_best_avg_save = None
if (not train) or ((np.asarray([
ep_save, is_best_save, is_best_avg_save
]) == None).sum() == 3):
to_save = False
else:
model.to_cpu()
saver.save_checkpoint(save_dict={
'model_params':
[entity.state_dict() for entity in model.entities]
},
episode=ep_save,
is_best=is_best_save,
is_best_avg=is_best_avg_save)
to_save = True
if (i_episode + 1) % 100 == 0:
summary = summaries[0] if train else summaries[1]
summary.update_log(i_episode,
episode_rewards.sum(),
list(episode_rewards.reshape(-1, )),
critic_loss=critic_losses,
actor_loss=actor_losses,
to_save=to_save,
comm_reward_total=episode_aux_rewards.sum(),
comm_reward_agents=list(
episode_aux_rewards.reshape(-1, )))
# Save gif
dir_monitor = config['dir_monitor_train'] if train else config[
'dir_monitor_test']
if config['render'] > 0 and i_episode % config['render'] == 0:
if config['env_id'] == 'waterworld':
imageio.mimsave('{}/{}.gif'.format(dir_monitor, i_episode),
frames[0::3])
else:
imageio.mimsave('{}/{}.gif'.format(dir_monitor, i_episode),
frames)
# <-- end loop: i_episode
if train:
print('Training completed')
else:
print('Test completed')
return (episode_rewards_all, episode_aux_rewards_all, episode_comm_all,
episode_comm_any_all)
def find_wl_contours2(im_ms, im_labels, cloud_mask, buffer_size, im_ref_buffer):
"""
New robust method for extracting shorelines. Incorporates the classification
component to refine the treshold and make it specific to the sand/water interface.
KV WRL 2018
Arguments:
-----------
im_ms: np.array
RGB + downsampled NIR and SWIR
im_labels: np.array
3D image containing a boolean image for each class in the order (sand, swash, water)
cloud_mask: np.array
2D cloud mask with True where cloud pixels are
buffer_size: int
size of the buffer around the sandy beach over which the pixels are considered in the
thresholding algorithm.
im_ref_buffer: np.array
binary image containing a buffer around the reference shoreline
Returns:
-----------
contours_wi: list of np.arrays
contains the coordinates of the contour lines extracted from the
NDWI (Normalized Difference Water Index) image
contours_mwi: list of np.arrays
contains the coordinates of the contour lines extracted from the
MNDWI (Modified Normalized Difference Water Index) image
"""
nrows = cloud_mask.shape[0]
ncols = cloud_mask.shape[1]
# calculate Normalized Difference Modified Water Index (SWIR - G)
im_mwi = SDS_tools.nd_index(im_ms[:,:,4], im_ms[:,:,1], cloud_mask)
# calculate Normalized Difference Modified Water Index (NIR - G)
im_wi = SDS_tools.nd_index(im_ms[:,:,3], im_ms[:,:,1], cloud_mask)
# stack indices together
im_ind = np.stack((im_wi, im_mwi), axis=-1)
vec_ind = im_ind.reshape(nrows*ncols,2)
# reshape labels into vectors
vec_sand = im_labels[:,:,0].reshape(ncols*nrows)
vec_water = im_labels[:,:,2].reshape(ncols*nrows)
# create a buffer around the sandy beach
se = morphology.disk(buffer_size)
im_buffer = morphology.binary_dilation(im_labels[:,:,0], se)
vec_buffer = im_buffer.reshape(nrows*ncols)
# select water/sand/swash pixels that are within the buffer
int_water = vec_ind[np.logical_and(vec_buffer,vec_water),:]
int_sand = vec_ind[np.logical_and(vec_buffer,vec_sand),:]
# make sure both classes have the same number of pixels before thresholding
if len(int_water) > 0 and len(int_sand) > 0:
if np.argmin([int_sand.shape[0],int_water.shape[0]]) == 1:
int_sand = int_sand[np.random.choice(int_sand.shape[0],int_water.shape[0], replace=False),:]
else:
int_water = int_water[np.random.choice(int_water.shape[0],int_sand.shape[0], replace=False),:]
# threshold the sand/water intensities
int_all = np.append(int_water,int_sand, axis=0)
t_mwi = filters.threshold_otsu(int_all[:,0])
t_wi = filters.threshold_otsu(int_all[:,1])
# find contour with MS algorithm
im_wi_buffer = np.copy(im_wi)
im_wi_buffer[~im_ref_buffer] = np.nan
im_mwi_buffer = np.copy(im_mwi)
im_mwi_buffer[~im_ref_buffer] = np.nan
contours_wi = measure.find_contours(im_wi_buffer, t_wi)
contours_mwi = measure.find_contours(im_mwi_buffer, t_mwi)
# remove contour points that are NaNs (around clouds)
contours = contours_wi
contours_nonans = []
for k in range(len(contours)):
if np.any(np.isnan(contours[k])):
index_nan = np.where(np.isnan(contours[k]))[0]
contours_temp = np.delete(contours[k], index_nan, axis=0)
if len(contours_temp) > 1:
contours_nonans.append(contours_temp)
else:
contours_nonans.append(contours[k])
contours_wi = contours_nonans
# repeat for MNDWI contours
contours = contours_mwi
contours_nonans = []
for k in range(len(contours)):
if np.any(np.isnan(contours[k])):
index_nan = np.where(np.isnan(contours[k]))[0]
contours_temp = np.delete(contours[k], index_nan, axis=0)
if len(contours_temp) > 1:
contours_nonans.append(contours_temp)
else:
contours_nonans.append(contours[k])
contours_mwi = contours_nonans
return contours_wi, contours_mwi
def calc_templatematch(Img, template = []):
return match_template(Img, template)[0,0] if np.any(template) else 0
def destruct(image,
mask,
imagesz=(210, 238, 200),
inputsz=(1, 32, 32, 32),
outputsz=(2, 14, 14, 14)):
"""The input is an image and a mask of shape (modalities, imagesz) and (imagesz).
The method decompses it into patches of size (inputsz) and (outputsz[1-3]).
Finally it return a numpy array of shape (number of patches, inputsz) and
(number of patches, 1, outputsz[1-3])"""
cut_mask = np.zeros(imagesz)
while not np.any(cut_mask):
index = random.randrange(imagesz[0])
cut_img = image[:, index]
cut_mask = mask[index]
n_height = ceil(imagesz[0] / outputsz[1])
n_width = ceil(imagesz[1] / outputsz[2])
n_depth = ceil(imagesz[2] / outputsz[3])
n_patches = n_height * n_width * n_depth
patches_mask = np.zeros((n_patches, outputsz[1], outputsz[2], outputsz[3]))
patches_image = np.zeros(
(n_patches, inputsz[0], inputsz[1], inputsz[2], inputsz[3]))
big_cut_mask = np.zeros(
(imagesz[0] + inputsz[1] * 2, imagesz[1] + inputsz[2] * 2,
imagesz[2] + inputsz[3] * 2))
big_cut_mask[inputsz[1]:inputsz[1] + imagesz[0],
inputsz[2]:inputsz[2] + imagesz[1],
inputsz[3]:inputsz[3] + imagesz[2]] = mask
big_cut_img = np.zeros(
(inputsz[0], imagesz[0] + inputsz[1] * 2, imagesz[1] + inputsz[2] * 2,
imagesz[2] + inputsz[3] * 2))
big_cut_img[:, inputsz[1]:inputsz[1] + imagesz[0],
inputsz[2]:inputsz[2] + imagesz[1],
inputsz[3]:inputsz[3] + imagesz[2]] = image
for x in range(n_height):
for y in range(n_width):
for z in range(n_depth):
j = x * outputsz[1]
k = y * outputsz[2]
l = z * outputsz[3]
shift_j = int((inputsz[1] - outputsz[1]) / 2)
shift_k = int((inputsz[2] - outputsz[2]) / 2)
shift_l = int((inputsz[3] - outputsz[3]) / 2)
patches_mask[x * n_width * n_depth + y * n_depth +
z] = big_cut_mask[inputsz[1] + j:inputsz[1] + j +
outputsz[1], inputsz[2] +
k:inputsz[2] + k + outputsz[2],
inputsz[3] + l:inputsz[3] + l +
outputsz[3]]
patches_image[x * n_width * n_depth + y * n_depth +
z] = big_cut_img[:, inputsz[1] + j -
shift_j:inputsz[1] + j +
inputsz[1] - shift_j,
inputsz[2] + k -
shift_k:inputsz[2] + k +
inputsz[2] - shift_k,
inputsz[3] + l -
shift_l:inputsz[3] + l +
inputsz[3] - shift_l]
patches_mask = np.reshape(
patches_mask, (n_patches, 1, outputsz[1], outputsz[2], outputsz[3]))
return patches_image, patches_mask
def fmt_export(arr, delimiter='\t', header=True, sig_fig=8, width='auto', justify='left', sign=False, pad=''):
"""
Create a format string for array `arr` such that columns are aligned in the output file when
saving with np.savetxt
"""
with np.testing.suppress_warnings() as sup:
sup.filter(RuntimeWarning)
flag1 = '' if justify != 'left' else '-'
flag2 = '+' if sign else ''
flag3 = '0' if pad == '0' else ''
fmt = []
hdr = []
for j, name in enumerate(arr.dtype.names):
dtype = arr[name].dtype
if dtype.kind in ['b']:
specifier = 'i'
precision = ''
w = 4 if np.all(arr[name]) else 5
elif dtype.kind in ['i', 'u']:
specifier = 'i'
precision = ''
w = _get_f_width(arr[name], sign)
elif dtype.kind in ['f', 'c']:
specifier = 'g'
precision = '.' + str(sig_fig)
# float notation width
# todo check for nan widths
w_f = _get_f_width(arr[name], sign) + sig_fig
# scientific notation width
i = 1 if sign or np.any(arr[name] < 0) else 0
w_s = sig_fig + 4 + i + 1 # +1 for decimal point which is not always needed
w = min(w_f, w_s) + 1
elif dtype.kind in ['U', 'S', 'O']:
specifier = 's'
precision = ''
w = np.max([len(str(item)) for item in arr[name]])
else:
raise TypeError(f'Invalid dtype kind {dtype.kind} for field {name}')
if width == 'auto':
col_w = w
elif isinstance(width, int):
col_w = width
else:
raise ValueError('Invalid width')
if header:
i = 2 if j == 0 else 0 # Additional space for header comment #
if width == 'auto':
_width = max(col_w, len(name) + i)
elif isinstance(width, int):
_width = col_w
func = str.ljust if justify == 'left' else str.rjust
fill = flag3 if flag3 else ' '
h = func(name, _width - i, fill)
hdr.append(h)
else:
_width = col_w
s = f'%{flag1}{flag2}{flag3}{_width}{precision}{specifier}'
fmt.append(s)
fmt = delimiter.join(fmt)
hdr = delimiter.join(hdr)
return fmt, hdr
def runme(id=None,exclude=None,benchmark='nightly',procedure='check',output='none',rank=1,numprocs=1):
"""
RUNME - test deck for ISSM nightly runs
In a test deck directory (tests/Vertification/NightlyRun for example)
The following command will launch all the existing tests:
>> runme()
To run the tests 101 and 102:
>> runme(id=[101,102])
etc...
Available options:
'id' followed by the list of ids requested
'exclude' ids to be excluded from the test
'benchmark' 'all' (all of the tests)
'nightly' (nightly run/ daily run)
'ismip' : validation of ismip-hom tests
'eismint': validation of eismint tests
'thermal': validation of thermal tests
'mesh' : validation of mesh tests
'adolc' : validation of adolc tests
'slr' : validation of slr tests
'procedure' 'check' : run the test (default)
'update': update the archive
Usage:
runme(varargin)
Examples:
runme()
runme(exclude=101)
runme(id=102,procedure='update')
"""
from parallelrange import parallelrange
from IdToName import IdToName
from arch import archread
from arch import archwrite
from arch import archdisp
#Get ISSM_DIR variable
ISSM_DIR=os.environ['ISSM_DIR']
#Process options
#GET benchmark {{{
if not benchmark in ['all','nightly','ismip','eismint','thermal','mesh','validation','tranforcing','adolc','slr','referential']:
print("runme warning: benchmark '{}' not supported, defaulting to test 'nightly'.".format(benchmark))
benchmark='nightly'
# }}}
#GET procedure {{{
if not procedure in ['check','update']:
print("runme warning: procedure '{}' not supported, defaulting to test 'check'.".format(procedure))
procedure='check'
# }}}
#GET output {{{
if not output in ['nightly','none']:
print("runme warning: output '{}' not supported, defaulting to test 'none'.".format(output))
output='none'
# }}}
#GET RANK and NUMPROCS for multithreaded runs {{{
if (numprocs<rank):
numprocs=1
# }}}
#GET ids {{{
flist=glob('test*.py') #File name must start with 'test' and must end by '.py' and must be different than 'test.py'
list_ids=[int(file[4:-3]) for file in flist if not file == 'test.py'] #Keep test id only (skip 'test' and '.py')
#print 'list_ids =',list_ids
i1,i2=parallelrange(rank,numprocs,len(list_ids)) #Get tests for this cpu only
list_ids=list_ids[i1:i2+1]
#print 'list_ids after parallelrange =',list_ids
if id:
if isinstance(id,list):
test_ids=id
else:
test_ids=[id]
test_ids=set(test_ids).intersection(set(list_ids))
else:
test_ids=set(list_ids)
#print 'test_ids after list =',test_ids
# }}}
#GET exclude {{{
if exclude:
if isinstance(exclude,list):
exclude_ids=exclude
else:
exclude_ids=[exclude]
test_ids=test_ids.difference(set(exclude_ids))
# print 'test_ids after exclude =',test_ids
# }}}
#Process Ids according to benchmarks {{{
if benchmark=='nightly':
test_ids=test_ids.intersection(set(range(1,1000)))
elif benchmark=='validation':
test_ids=test_ids.intersection(set(range(1001,2000)))
elif benchmark=='ismip':
test_ids=test_ids.intersection(set(range(1101,1200)))
elif benchmark=='eismint':
test_ids=test_ids.intersection(set(range(1201,1300)))
elif benchmark=='thermal':
test_ids=test_ids.intersection(set(range(1301,1400)))
elif benchmark=='mesh':
test_ids=test_ids.intersection(set(range(1401,1500)))
elif benchmark=='tranforcing':
test_ids=test_ids.intersection(set(range(1501,1503)))
elif benchmark=='referential':
test_ids=test_ids.intersection(set(range(1601,1603)))
elif benchmark=='slr':
test_ids=test_ids.intersection(set(range(2001,2500)))
elif benchmark=='adolc':
test_ids=test_ids.intersection(set(range(3001,3200)))
#print 'test_ids after benchmark =',test_ids
test_ids=list(test_ids)
test_ids.sort()
#print 'test_ids after sort =',test_ids
# }}}
#Loop over tests and launch sequence
root=os.getcwd()
for id in test_ids:
print "----------------starting:%i-----------------------" % id
try:
#Execute test
os.chdir(root)
id_string=IdToName(id)
execfile('test'+str(id)+'.py',globals())
#UPDATE ARCHIVE?
archive_name='Archive'+str(id)
if procedure=='update':
archive_file=os.path.join('..','Archives',archive_name+'.arch')
if os.path.isfile(archive_file):
os.remove(archive_file)
for k,fieldname in enumerate(field_names):
field=np.array(field_values[k],dtype=float)
if len(field.shape) == 1:
if np.size(field):
field=field.reshape(np.size(field),1)
else:
field=field.reshape(0,0)
elif len(field.shape) == 0:
field=field.reshape(1,1)
# Matlab uses base 1, so use base 1 in labels
archwrite(archive_file,archive_name+'_field'+str(k+1),field)
print "File '%s' saved.\n" % os.path.join('..','Archives',archive_name+'.arch')
#ELSE: CHECK TEST
else:
#load archive
if os.path.exists(os.path.join('..','Archives',archive_name+'.arch')):
archive_file=os.path.join('..','Archives',archive_name+'.arch')
else:
raise IOError("Archive file '"+os.path.join('..','Archives',archive_name+'.arch')+"' does not exist.")
for k,fieldname in enumerate(field_names):
try:
#Get field and tolerance
field=np.array(field_values[k])
if len(field.shape) == 1:
if np.size(field):
field=field.reshape(np.size(field),1)
else:
field=field.reshape(0,0)
tolerance=field_tolerances[k]
#compare to archive
# Matlab uses base 1, so use base 1 in labels
archive=np.array(archread(archive_file,archive_name+'_field'+str(k+1)))
if archive == None:
raise NameError("Field name '"+archive_name+'_field'+str(k+1)+"' does not exist in archive file.")
error_diff=np.amax(np.abs(archive-field),axis=0)/(np.amax(np.abs(archive),axis=0)+float_info.epsilon)
#disp test result
if (np.any(error_diff>tolerance) or np.isnan(error_diff)):
print('ERROR difference: {} > {} test id: {} test name: {} field: {}'.format(error_diff,tolerance,id,id_string,fieldname))
else:
print('SUCCESS difference: {} < {} test id: {} test name: {} field: {}'.format(error_diff,tolerance,id,id_string,fieldname))
except Exception as message:
#something went wrong, print failure message:
print format_exc()
directory=os.getcwd().split('/') # not used?
if output=='nightly':
fid=open(os.path.join(ISSM_DIR,'nightlylog','pythonerror.log'), 'a')
fid.write('%s' % message)
fid.write('\n------------------------------------------------------------------\n')
fid.close()
print('FAILURE difference: N/A test id: {} test name: {} field: {}'.format(id,id_string,fieldname))
else:
print('FAILURE difference: N/A test id: {} test name: {} field: {}'.format(id,id_string,fieldname))
raise RuntimeError(message)
except Exception as message:
#something went wrong, print failure message:
print format_exc()
directory=os.getcwd().split('/') # not used?
if output=='nightly':
fid=open(os.path.join(ISSM_DIR,'nightlylog','pythonerror.log'), 'a')
fid.write('%s' % message)
fid.write('\n------------------------------------------------------------------\n')
fid.close()
print('FAILURE difference: N/A test id: {} test name: {} field: {}'.format(id,id_string,'N/A'))
else:
print('FAILURE difference: N/A test id: {} test name: {} field: {}'.format(id,id_string,'N/A'))
raise RuntimeError(message)
print "----------------finished:%i-----------------------" % id
return
def join_bins(bin_edges, probabilities, min_prob=0.05):
"""Join bins until at least the minimum probability is contained.
By joining adjacent bins, find the configuration with the maximum
number of bins that each contain at least a certain probability.
Parameters
----------
bin_edges : iterable of 2-tuple
Upper and lower bounds associated with each bin.
probabilities : array-like
Probabilities in each bin. The contiguous ranges where bin joining
must be attempted will automatically be determined.
min_prob : float
The minimum probability (inclusive) per (final) bin.
Returns
-------
bin_edges : list of 2-tuple
List containing the new bin edges.
probabilities : array
Probabilities corresponding to the above bins.
Raises
------
ValueError : If the number of bin edges does not match the number
of probabilities.
ValueError : If the sum of all `probabilities` does not exceed `min_prob`.
"""
if len(bin_edges) != len(probabilities):
raise ValueError("Length of bin_edges and probabilities must match.")
probabilities = np.asarray(probabilities)
if np.sum(probabilities) <= min_prob:
raise ValueError("Sum of probabilities must exceed min_prob.")
max_i = probabilities.shape[0] - 1
def _join(start_i, end_i):
"""Return new bin edges and probabilities after a join.
Parameters
----------
start_i : int
Beginning of the join.
end_i : int
End of the join.
Returns
-------
bin_edges : list of 2-tuple
List containing the new bin edges.
probabilities : array
Probabilities corresponding to the above bins.
"""
new_bin_edges = []
new_probabilities = []
# Remove all but the first index within the join window.
for i in filterfalse(
lambda x: x in range(start_i + 1, end_i + 1), range(max_i + 1),
):
if i == start_i:
new_bin_edges.append((bin_edges[start_i][0], bin_edges[end_i][1],))
new_probabilities.append(np.sum(probabilities[start_i : end_i + 1]))
else:
new_bin_edges.append(bin_edges[i])
new_probabilities.append(probabilities[i])
return join_bins(
bin_edges=new_bin_edges, probabilities=new_probabilities, min_prob=min_prob,
)
# Identify regions with low probabilities.
join_mask = probabilities < min_prob
if not np.any(join_mask):
# Joining is complete.
return (bin_edges, probabilities)
# Find the contiguous clusters.
labelled, n_clusters = label(join_mask)
variations = []
# Carry out contiguous bin joining around all clusters.
for cluster_i in range(1, n_clusters + 1):
cluster_indices = np.where(labelled == cluster_i)[0]
cluster_bounds = (cluster_indices[0], cluster_indices[-1])
# Also consider the adjacent bins, since this may be needed
# in some cases.
join_bounds = tuple(
np.clip(np.array([cluster_bounds[0] - 1, cluster_bounds[1] + 1]), 0, max_i)
)
for start_i in range(*join_bounds):
# Optimisation: prevent 'orphan' bins on the left.
if join_bounds[0] == 0 and start_i != 0:
continue
for end_i in range(start_i + 1, join_bounds[1] + 1):
# Optimisation: prevent 'orphan' bins on the right.
if join_bounds[1] == max_i and end_i != max_i:
continue
# If the sum of probabilities between `start_i` and `end_i`
# exceeds the minimum threshold, join the bins.
if np.sum(probabilities[start_i : end_i + 1]) >= min_prob:
variations.append(_join(start_i, end_i))
# Return the 'best' variation - the set of bins with the lowest variability,
# measured using the standard deviation of the probabilities. Only sets with
# the largest number of bins will be considered here.
lengths = [len(variation[0]) for variation in variations]
max_length = max(lengths)
long_variations = [
variation for variation in variations if len(variation[0]) == max_length
]
variation_stds = [np.std(variation[1]) for variation in long_variations]
min_index = np.argsort(variation_stds)[0]
return long_variations[min_index]
def draw_static(self, axis=plt.gca(), use_data=True, filetypes=['eps', 'pdf', 'png'], save=False):
self.axis = axis
#The orbital view. Draw points, subsolar longitude line, and labels.
P = self.planet.P
e = self.planet.e
a = (self.planet.a / U.AU).value
#If use_data is True, the dots showing the equal time intervals over the orbit will be colored according to where the data in each band exist. Note: this works best when the time sampling of the data is finer than the plotting time resolution.
if use_data:
coverage = N.zeros(self.time_resolution+1, dtype=bool)
partial_phase = True
t_set = {}
for i, band in enumerate(['3p6','4p5','8p0']):
if band in self.data:
t_set[band] = N.unique(self.data[band]['t'])
data = t_set[band]
band_range = N.array([N.any(N.abs(t-data) <= 0.5*P/U.d/self.time_resolution) for t in self.times/U.d])
coverage = coverage | band_range
#The "stitch" is that the array for checking coverage has its first element overlapping with the last. So if either are covered, both should be covered.
coverage_stitch = coverage[0] | coverage[-1]
coverage[0] = coverage_stitch
coverage[-1] = coverage_stitch
#Boolean mask for whether we consider the photometry a partial, rather than full, phase curve. I chose to call any light curve with a time range < 90% of the orbital period as "partial".
partial_phase = partial_phase & ((N.max(t_set[band])-N.min(t_set[band]))*U.d/P < 0.8)
orbit_range = axis.scatter(self.x_points[band_range], self.y_points[band_range], color=color_datlab[band], s=(6*(i+1)**2), zorder=(3-i)*5)
orbit_gap = axis.scatter(self.x_points[~coverage], self.y_points[~coverage], color='k', s=2, zorder=100)
else:
orbit_points = axis.scatter(self.x_points, self.y_points, color='k', s=2, zorder=1)
peri_angle = ((self.planet.w + 180*U.deg)%(360*U.deg)).to(U.deg)
orbit_outline = self.axis.add_artist(Ellipse(xy = (a*e*N.cos(peri_angle+180*U.deg), a*e*N.sin(peri_angle+180*U.deg)),
#orbit_outline = self.axis.add_artist(Ellipse(xy = (a*e*N.cos(peri_angle+0*U.deg), a*e*N.sin(peri_angle+0*U.deg)),
width = 2*a,
height = 2*a*N.sqrt(1-e**2),
angle = peri_angle.value,
fill = False, edgecolor = 'k', alpha = 0.3, linestyle = (0, (0.5, 1.5)) ))
#The star point scales with the radius of the actual star, and is set true to the scale of the plotted orbit semimajor axis.
star_scale = float(self.planet.R/U.AU)
#The color is calculated from the blackbody color at the effective temperature.
star_color = cm.irgb_string_from_xyz(bb.blackbody_color(float(self.planet.Teff/U.K)))
limb_darkening_span = N.squeeze(N.array([colors.to_rgba_array(cm.irgb_string_from_xyz(bb.blackbody_color(T))) for T in N.linspace(0.8, 1, num=20)*self.planet.Teff/U.K]))
#star_point = axis.scatter(0, 0, color=star_color, s=star_scale, edgecolors='k')
for n, shade in enumerate(limb_darkening_span[:,:-1]):
self.star_point = axis.add_artist(Ellipse(xy=[0,0], width=(star_scale*2)*(1-n/20), height=(star_scale*2)*(1-n/20), angle=0))
self.star_point.set_facecolor(shade)
#Draw the line from star to apastron, and the line to periastron.
if e > 0:
x_apo = N.cos(self.planet.w) * N.array([star_scale, self.orbit_scale*(1+e)])
y_apo = N.sin(self.planet.w) * N.array([star_scale, self.orbit_scale*(1+e)])
apo_line = axis.plot(x_apo, y_apo, color='#444444', linestyle='-', linewidth=1, dashes=[2,4])
x_peri = N.cos(180*U.deg+self.planet.w) * N.array([star_scale, self.orbit_scale*(1-e)])
y_peri = N.sin(180*U.deg+self.planet.w) * N.array([star_scale, self.orbit_scale*(1-e)])
peri_line = axis.plot(x_peri, y_peri, color='#444444', linestyle='-', linewidth=1, dashes=[1,2])
line_of_nodes = axis.axhline(y=0, linewidth=0.5, color='k', alpha=0.5, linestyle=(0, (0.25, 1.75)), zorder=1)
#Draw the band representing the geometric transit and eclipse window.
occultation_band = axis.axvspan(xmin=-star_scale, xmax=star_scale, facecolor='0.2', alpha=0.25)
#Draw some motion arrows following a slightly larger elliptical arc around the orbit points.
offset_factor = 0.1
motion_arcs = [ Arc(xy = (a*e*N.cos(peri_angle+180*U.deg), a*e*N.sin(peri_angle+180*U.deg)),
width = (1+offset_factor)*(2*a),
height = (1+offset_factor)*(2*a*N.sqrt(1-e**2)),
angle = peri_angle.value,
theta1 = i,
theta2 = j ) for (i,j) in [(-20,-10), (160, 170)] ]
motion_paths = [(motion_arc.get_transform()).transform_path(motion_arc.get_path()) for motion_arc in motion_arcs]
motion_lines = [axis.add_artist(FancyArrowPatch(path=motion_path, arrowstyle='->', mutation_scale=10, color='#444444')) for motion_path in motion_paths]
#Remove axis labels and ticks.
plt.setp(plt.getp(axis.axes, 'xticklabels'), visible=False)
plt.setp(plt.getp(axis.axes, 'yticklabels'), visible=False)
axis.set_aspect('equal')
#Tight boundaries for the figure, using the span of the orbit.
if use_data:
if partial_phase:
orbit_spanx = [N.min(self.x_points[coverage]) - 0.1*(N.max(self.x_points[coverage])-N.min(self.x_points[coverage])), N.max(self.x_points[coverage]) + 0.1*(N.max(self.x_points[coverage])-N.min(self.x_points[coverage]))]
orbit_spany = [N.min(self.y_points[coverage]) - 0.1*(N.max(self.y_points[coverage])-N.min(self.y_points[coverage])), N.max(self.y_points[coverage]) + 0.1*(N.max(self.y_points[coverage])-N.min(self.y_points[coverage]))]
axis.set_xlim(orbit_spanx)
axis.set_ylim(orbit_spany)
else:
orbit_spanx = [N.min(self.x_points) - 0.1*(N.max(self.x_points)-N.min(self.x_points)), N.max(self.x_points) + 0.1*(N.max(self.x_points)-N.min(self.x_points))]
orbit_spany = [N.min(self.y_points) - 0.1*(N.max(self.y_points)-N.min(self.y_points)), N.max(self.y_points) + 0.1*(N.max(self.y_points)-N.min(self.y_points))]
axis.set_xlim(orbit_spanx)
axis.set_ylim(orbit_spany)
#The observer line points from the star to the bottom of the figure, which represents the direction to the observer.
observer_linelength = self.orbit_scale * (1-e**2) / (1 + e*N.cos(self.planet.w+90*U.deg))
xmin, xmax = axis.get_xlim()
ymin, ymax = axis.get_ylim()
observer_line = axis.arrow(xmin+0.05*(xmax-xmin), ymin+0.16*(ymax-ymin), 0, -0.05*(ymax-ymin), color='#444444', head_width=0.015*(ymax-ymin), linewidth=2, width=0.001*(ymax-ymin))
earth_label = axis.text(xmin+0.05*(xmax-xmin), ymin+0.08*(ymax-ymin),'$\mathbf{\oplus}$', horizontalalignment='center', verticalalignment='top', fontproperties=fontprops)
if e > 0:
PSR_x = self.x_points[1:-1]
PSR_y = self.y_points[1:-1]
PSR_lines = [axis.add_artist(FancyArrow(x, y, 0.03*(ymax-ymin)*N.cos(PSR_line), 0.03*(ymax-ymin)*N.sin(PSR_line), color='black', lw=0.25, head_length=0.001, shape='right')) for (x, y, PSR_line) in zip(PSR_x, PSR_y, self.PSR_angle[1:-1])]
subsolar_lines = [axis.plot([x, x+0.045*(ymax-ymin)*N.cos(subsolar_line)], [y, y+0.045*(ymax-ymin)*N.sin(subsolar_line)], color='#4E89B6', lw=1) for (x, y, subsolar_line) in zip(self.x_points[:-1], self.y_points[:-1], self.subsolar_angle[:-1])]
plt.tick_params(
axis='x', # changes apply to the x-axis
which='both', # both major and minor ticks are affected
bottom='off', # ticks along the bottom edge are off
top='off', # ticks along the top edge are off
labelbottom='off') # labels along the bottom edge are off
plt.tick_params(
axis='y', # changes apply to the x-axis
which='both', # both major and minor ticks are affected
left='off', # ticks along the bottom edge are off
right='off', # ticks along the top edge are off
labelleft='off') # labels along the bottom edge are off
#Instead of explicit axes we opt for a scale bar, here chosen to be 0.01 AU.
from mpl_toolkits.axes_grid1.anchored_artists import AnchoredSizeBar
scalebar = AnchoredSizeBar(axis.transData,
0.01,
r"0.01 AU",
loc=1,
pad=0.0, borderpad=0.5, sep=5,
frameon=False,
fontproperties=fontprops)
axis.add_artist(scalebar)
#If we're plotting the whole orbit, the bounding box for the text showing the orbital period should be able to find room in the lower-right corner. Otherwise, there is a chance something could intersect it, and so we draw a white bounding box around it.
if partial_phase: textbox_settings = dict(pad=0, fc='white', ec='white')
else: textbox_settings = dict(pad=0, alpha=0)
axis.text(0.975, 0.05,'$P = $ {0:.2f}'.format(self.planet.P), horizontalalignment='right',
verticalalignment='center',
transform=axis.transAxes,
fontproperties=fontprops,
bbox = textbox_settings)
if save:
for filetype in filetypes:
plt.savefig('{0}_{1}_orbit.{2}'.format(str(datetime.date.today()).replace(" ", ""), (self.planet.name).replace(" ", ""), filetype), bbox_inches='tight', transparent=True)
def add_recommendations(self, is_relevant, user_id):
self.users_mask[user_id] = np.any(is_relevant)
def __preprocess_individual(file, outdir, overwrite):
""" Internal function for preprocessing raw MEG files
:param file:
input file along with path
:param outdir:
where we want to save this
:return: save_file_path:
a path to the saved and filtered file
"""
save_file_path = ""
no_ecg_removed = False
no_eog_removed = False
num = os.path.basename(file).split('_')[0]
if os.path.basename(file).split('raw')[1] != '.fif':
append = os.path.basename(file).split('raw')[1].split('.fif')[0]
else:
append = ''
append = 'raw' + append + '.fif'
base = os.path.basename(file).split('.')[0]
# check if any of these files exists, if not overwrite then skip and return path
# could be any of these names
check_fnames = [
f'{outdir}/{num}_noeog_noecg_clean_{append}',
f'{outdir}/{num}_noecg_clean_{append}',
f'{outdir}/{num}_noeog_clean_{append}',
f'{outdir}/{num}_clean_{append}'
]
if np.any([os.path.isfile(f) for f in check_fnames]):
index = np.where([os.path.isfile(f) for f in check_fnames])[0]
if not overwrite:
print(
f'file for {os.path.basename(file)} already exists, skipping to next'
)
save_file_path = check_fnames[index[0]]
return save_file_path
# read file
try:
raw = mne.io.read_raw_fif(file, preload=True)
except OSError:
print('could not read ' + file)
return ''
raw.filter(0.1, None)
raw.notch_filter(np.arange(50, 241, 50),
filter_length='auto',
phase='zero')
# Run ICA on raw data to find blinks and eog
try:
ica = mne.preprocessing.ICA(n_components=25, method='picard').fit(raw)
except:
raw.crop(1) # remove the first second due to NaNs
ica = mne.preprocessing.ICA(n_components=25, method='picard').fit(raw)
try:
# look for and remove EOG
eog_epochs = mne.preprocessing.create_eog_epochs(
raw) # get epochs of eog (if this exists)
eog_inds, eog_scores = ica.find_bads_eog(
eog_epochs) # try and find correlated components
# define flags for tracking if we found components matching or not
no_ecg_removed = False
no_eog_removed = False
# if we have identified something !
if len(eog_inds) > 0:
ica.exclude.extend(eog_inds)
else:
print(
f'{os.path.basename(file)} cannot detect eog automatically manual ICA must be done'
)
no_eog_removed = True
except:
print(
f'{os.path.basename(file)} cannot detect eog automatically manual ICA must be done'
)
no_eog_removed = True
# now we do this with hearbeat
try:
ecg_epochs = mne.preprocessing.create_ecg_epochs(
raw) # get epochs of eog (if this exists)
ecg_inds, ecg_scores = ica.find_bads_ecg(
ecg_epochs) # try and find correlated components
# if one component reaches above threshold then remove components automagically
if len(ecg_inds) > 0:
ica.exclude.extend(ecg_inds) # exclude top 3 components
else: # flag for manual ICA inspection and removal
print(
f'{os.path.basename(file)} cannot detect ecg automatically manual ICA must be done'
)
no_ecg_removed = True
except:
print(
f'{os.path.basename(file)} cannot detect ecg automatically manual ICA must be done'
)
no_ecg_removed = True
# exclude all labelled components
raw = ica.apply(inst=raw)
# save the file
if no_ecg_removed and no_eog_removed:
outfname = f'{outdir}/{num}_noeog_noecg_clean_{append}'
elif no_ecg_removed:
outfname = f'{outdir}/{num}_noecg_clean_{append}'
elif no_eog_removed:
outfname = f'{outdir}/{num}_noeog_clean_{append}'
else:
outfname = f'{outdir}/{num}_clean_{append}'
raw = raw.resample(250)
raw.save(outfname, overwrite=overwrite)
save_file_path = outfname
# return
return save_file_path
def mass_conservation_inversion(gdir,
glen_a=None,
fs=None,
write=True,
filesuffix=''):
""" Compute the glacier thickness along the flowlines
More or less following Farinotti et al., (2009).
Parameters
----------
gdir : :py:class:`oggm.GlacierDirectory`
the glacier directory to process
glen_a : float
glen's creep parameter A
fs : float
sliding parameter
write: bool
default behavior is to compute the thickness and write the
results in the pickle. Set to False in order to spare time
during calibration.
filesuffix : str
add a suffix to the output file
"""
# Defaults
if glen_a is None:
glen_a = cfg.PARAMS['inversion_glen_a']
if fs is None:
fs = cfg.PARAMS['inversion_fs']
# Check input
if fs == 0.:
_inv_function = _inversion_simple
else:
_inv_function = _inversion_poly
# Ice flow params
fd = 2. / (cfg.PARAMS['glen_n'] + 2) * glen_a
a3 = fs / fd
rho = cfg.PARAMS['ice_density']
# Shape factor params
sf_func = None
# Use .get to obatin default None for non-existing key
# necessary to pass some tests
# TODO: remove after tests are adapted
use_sf = cfg.PARAMS.get('use_shape_factor_for_inversion')
if use_sf == 'Adhikari' or use_sf == 'Nye':
sf_func = utils.shape_factor_adhikari
elif use_sf == 'Huss':
sf_func = utils.shape_factor_huss
sf_tol = 1e-2 # TODO: better as params in cfg?
max_sf_iter = 20
# Clip the slope, in degrees
clip_angle = cfg.PARAMS['min_slope']
out_volume = 0.
cls = gdir.read_pickle('inversion_input')
for cl in cls:
# Clip slope to avoid negative and small slopes
slope = cl['slope_angle']
slope = np.clip(slope, np.deg2rad(clip_angle), np.pi / 2.)
# Parabolic bed rock
w = cl['width']
a0s = -cl['flux_a0'] / ((rho * cfg.G * slope)**3 * fd)
sf = np.ones(slope.shape) # Default shape factor is 1
# TODO: maybe take height update as criterion for iteration end instead
# of sf_diff?
if sf_func is not None:
# Start iteration for shape factor with guess of 1
i = 0
sf_diff = np.ones(slope.shape)
while i < max_sf_iter and np.any(sf_diff > sf_tol):
out_thick = _compute_thick(gdir, a0s, a3, cl['flux_a0'], sf,
_inv_function)
sf_diff[:] = sf[:]
sf = sf_func(w, out_thick, cl['is_rectangular'])
sf_diff = sf_diff - sf
i += 1
# TODO: Iteration at the moment for all grid points,
# even if some already converged. Change?
log.info('Shape factor {:s} used, took {:d} iterations for '
'convergence.'.format(use_sf, i))
out_thick = _compute_thick(gdir, a0s, a3, cl['flux_a0'], sf,
_inv_function)
# volume
fac = np.where(cl['is_rectangular'], 1, 2. / 3.)
volume = fac * out_thick * w * cl['dx']
if write:
cl['thick'] = out_thick
cl['volume'] = volume
out_volume += np.sum(volume)
if write:
gdir.write_pickle(cls, 'inversion_output', filesuffix=filesuffix)
return out_volume, gdir.rgi_area_km2 * 1e6
def generateActivityDiagram(config,
shower_data,
ax_handle=None,
sol_marker=None):
""" Generates a plot of shower activity across all solar longitudes. """
shower_data = np.array(shower_data)
# Fill in min/max solar longitudes if they are not present
for shower in shower_data:
sol_min, sol_peak, sol_max = list(shower)[3:6]
if np.isnan(sol_min):
shower[3] = (sol_peak - config.shower_lasun_threshold) % 360
if np.isnan(sol_max):
shower[5] = (sol_peak + config.shower_lasun_threshold) % 360
# Sort showers by duration
durations = [(shower[5] - shower[3] + 180) % 360 - 180
for shower in shower_data]
shower_data = shower_data[np.argsort(durations)][::-1]
# Generate an array of shower activity per 1 deg of solar longitude
code_name_dict = {}
activity_stack = np.zeros((20, 360), dtype=np.uint16)
colors = cm.tab10(np.linspace(0, 1.0, 10))
shower_index = 0
for i in range(20):
for sol_plot in range(0, 360):
# If the cell is unassigned, check to see if a shower is active
if activity_stack[i, sol_plot] > 0:
continue
for shower in shower_data:
code = int(shower[0])
name = shower[1]
# Skip assigned showers
if code in code_name_dict:
continue
sol_min, sol_peak, sol_max = list(shower)[3:6]
sol_min = int(np.floor(sol_min)) % 360
sol_max = int(np.ceil(sol_max)) % 360
# If the shower is active at the given solar longitude and there aren't any other showers
# in the same activity period, assign shower code to this solar longitude
shower_active = False
if (sol_max - sol_min) < 180:
# Check if the shower is active
if (sol_plot >= sol_min) and (sol_plot <= sol_max):
# Leave a buffer of +/- 3 deg around the shower
sol_min_check = sol_min - 3
if sol_min_check < 0:
sol_min_check = 0
sol_max_check = sol_max + 3
if sol_max_check > 360:
sol_max_check = 360
# Check if the solar longitue range is free of other showers
if not np.any(
activity_stack[i,
sol_min_check:sol_max_check]):
# Assign the shower code to activity stack
activity_stack[i, sol_min:sol_max] = code
code_name_dict[code] = [name, sol_peak]
shower_active = True
else:
if (sol_plot >= sol_min) or (sol_plot <= sol_max):
# Check if the solar longitue range is free of other showers
if (not np.any(activity_stack[i, 0:sol_max])) and \
(not np.any(activity_stack[i, sol_min:])):
# Assign shower code to activity stack
activity_stack[i, 0:sol_max] = code
activity_stack[i, sol_min:] = code
shower_active = True
if shower_active:
# Get shower color
color = colors[shower_index % 10]
shower_index += 1
# Assign shower params
code_name_dict[code] = [
name, sol_min, sol_peak, sol_max, color
]
# If no axis was given, crate one
if ax_handle is None:
ax_handle = plt.subplot(111, facecolor='black')
# Set background color
plt.gcf().patch.set_facecolor('black')
# Change axis color
ax_handle.spines['bottom'].set_color('w')
ax_handle.spines['top'].set_color('w')
ax_handle.spines['right'].set_color('w')
ax_handle.spines['left'].set_color('w')
# Change tick color
ax_handle.tick_params(axis='x', colors='w')
ax_handle.tick_params(axis='y', colors='w')
# Change axis label color
ax_handle.yaxis.label.set_color('w')
ax_handle.xaxis.label.set_color('w')
# Plot the activity graph
active_shower = 0
vertical_scale_line = 0.5
vertical_shift_text = 0.01
text_size = 8
for i, line in enumerate(activity_stack):
for shower_block in line:
# If a new shower was found, plot it
if (shower_block != active_shower) and (shower_block > 0):
# Get shower parameters
name, sol_min, sol_peak, sol_max, color = code_name_dict[
shower_block]
# Plot the shower activity period
if (sol_max - sol_min) < 180:
x_arr = np.arange(sol_min, sol_max + 1, 1)
ax_handle.plot(x_arr, np.zeros_like(x_arr) + i*vertical_scale_line, linewidth=3, \
color=color, zorder=3)
ax_handle.text(round((sol_max + sol_min)/2), i*vertical_scale_line + vertical_shift_text, \
name, ha='center', va='bottom', color='w', size=text_size, zorder=3)
else:
x_arr = np.arange(0, sol_max + 1, 1)
ax_handle.plot(x_arr, np.zeros_like(x_arr) + i*vertical_scale_line, linewidth=3, \
color=color, zorder=3)
x_arr = np.arange(sol_min, 361, 1)
ax_handle.plot(x_arr, np.zeros_like(x_arr) + i*vertical_scale_line, linewidth=3, \
color=color, zorder=3)
ax_handle.text(0, i*vertical_scale_line + vertical_shift_text, name, ha='center', \
va='bottom', color='w', size=text_size, zorder=2)
# Plot peak location
ax_handle.scatter(sol_peak,
i * vertical_scale_line,
marker='+',
c="w",
zorder=4)
active_shower = shower_block
# Hide y axis
ax_handle.get_yaxis().set_visible(False)
ax_handle.set_xlabel('Solar longitude (deg)')
# Get the plot Y limits
y_min, y_max = ax_handle.get_ylim()
# Plot a line with given solver longitude
if sol_marker is not None:
# Plot the solar longitude line behind everything else
y_arr = np.linspace(y_min, y_max, 5)
if not isinstance(sol_marker, list):
sol_marker = [sol_marker]
for sol_value in sol_marker:
ax_handle.plot(np.zeros_like(y_arr) + sol_value, y_arr, color='r', linestyle='dashed', zorder=2, \
linewidth=1)
# Plot month names at the 1st of that month (start in April of this year)
for month_no, year_modifier in [[4, 0], [5, 0], [6, 0], [7, 0], [8, 0],
[9, 0], [10, 0], [11, 0], [12, 0], [1, 1],
[2, 1], [3, 1]]:
# Get the solar longitude of the 15th date of the month
curr_year = datetime.datetime.now().year
dt = datetime.datetime(curr_year + year_modifier, month_no, 15, 0, 0,
0)
sol = np.degrees(jd2SolLonSteyaert(datetime2JD(dt))) % 360
# Plot the month name in the background of the plot
plt.text(sol, y_max, dt.strftime("%b").upper(), alpha=0.3, rotation=90, size=20, \
zorder=1, color='w', va='top', ha='center')
# Get the solar longitude of the 1st date of the month
curr_year = datetime.datetime.now().year
dt = datetime.datetime(curr_year + year_modifier, month_no, 1, 0, 0, 0)
sol = np.degrees(jd2SolLonSteyaert(datetime2JD(dt))) % 360
# Plot the manth beginning line
y_arr = np.linspace(y_min, y_max, 5)
plt.plot(np.zeros_like(y_arr) + sol,
y_arr,
linestyle='dotted',
alpha=0.3,
zorder=3,
color='w')
# Force Y limits
ax_handle.set_ylim([y_min, y_max])
ax_handle.set_xlim([0, 360])
def add_output(self,
name,
val=1.0,
shape=None,
units=None,
res_units=None,
desc='',
lower=None,
upper=None,
ref=1.0,
ref0=0.0,
res_ref=1.0,
var_set=0):
"""
Add an output variable to the component.
Parameters
----------
name : str
name of the variable in this component's namespace.
val : float or list or tuple or ndarray
The initial value of the variable being added in user-defined units. Default is 1.0.
shape : int or tuple or list or None
Shape of this variable, only required if val is not an array.
Default is None.
units : str or None
Units in which the output variables will be provided to the component during execution.
Default is None, which means it has no units.
res_units : str or None
Units in which the residuals of this output will be given to the user when requested.
Default is None, which means it has no units.
desc : str
description of the variable.
lower : float or list or tuple or ndarray or Iterable or None
lower bound(s) in user-defined units. It can be (1) a float, (2) an array_like
consistent with the shape arg (if given), or (3) an array_like matching the shape of
val, if val is array_like. A value of None means this output has no lower bound.
Default is None.
upper : float or list or tuple or ndarray or or Iterable None
upper bound(s) in user-defined units. It can be (1) a float, (2) an array_like
consistent with the shape arg (if given), or (3) an array_like matching the shape of
val, if val is array_like. A value of None means this output has no upper bound.
Default is None.
ref : float or ndarray
Scaling parameter. The value in the user-defined units of this output variable when
the scaled value is 1. Default is 1.
ref0 : float or ndarray
Scaling parameter. The value in the user-defined units of this output variable when
the scaled value is 0. Default is 0.
res_ref : float or ndarray
Scaling parameter. The value in the user-defined res_units of this output's residual
when the scaled value is 1. Default is 1.
var_set : hashable object
For advanced users only. ID or color for this variable, relevant for reconfigurability.
Default is 0.
Returns
-------
dict
metadata for added variable
"""
if units == 'unitless':
warn_deprecation(
"Output '%s' has units='unitless' but 'unitless' "
"has been deprecated. Use "
"units=None instead. Note that connecting a "
"unitless variable to one with units is no longer "
"an error, but will issue a warning instead." % name)
units = None
# First, type check all arguments
if not isinstance(name, str):
raise TypeError('The name argument should be a string')
if not np.isscalar(val) and not isinstance(
val, (list, tuple, np.ndarray, Iterable)):
raise TypeError(
'The val argument should be a float, list, tuple, or ndarray')
if not np.isscalar(ref) and not isinstance(
val, (list, tuple, np.ndarray, Iterable)):
raise TypeError(
'The ref argument should be a float, list, tuple, or ndarray')
if not np.isscalar(ref0) and not isinstance(
val, (list, tuple, np.ndarray, Iterable)):
raise TypeError(
'The ref0 argument should be a float, list, tuple, or ndarray')
if not np.isscalar(res_ref) and not isinstance(
val, (list, tuple, np.ndarray, Iterable)):
raise TypeError(
'The res_ref argument should be a float, list, tuple, or ndarray'
)
if shape is not None and not isinstance(
shape, (int, tuple, list, np.integer)):
raise TypeError(
"The shape argument should be an int, tuple, or list but "
"a '%s' was given" % type(shape))
if units is not None and not isinstance(units, str):
raise TypeError('The units argument should be a str or None')
if res_units is not None and not isinstance(res_units, str):
raise TypeError('The res_units argument should be a str or None')
# Check that units are valid
if units is not None and not valid_units(units):
raise ValueError("The units '%s' are invalid" % units)
metadata = {}
# value, shape: based on args, making sure they are compatible
metadata['value'], metadata['shape'] = ensure_compatible(
name, val, shape)
metadata['size'] = np.prod(metadata['shape'])
# units, res_units: taken as is
metadata['units'] = units
metadata['res_units'] = res_units
# desc: taken as is
metadata['desc'] = desc
if lower is not None:
lower = ensure_compatible(name, lower, metadata['shape'])[0]
if upper is not None:
upper = ensure_compatible(name, upper, metadata['shape'])[0]
metadata['lower'] = lower
metadata['upper'] = upper
# All refs: check the shape if necessary
for item, item_name in zip([ref, ref0, res_ref],
['ref', 'ref0', 'res_ref']):
if not np.isscalar(item):
if np.atleast_1d(item).shape != metadata['shape']:
raise ValueError('The %s argument has the wrong shape' %
item_name)
if np.isscalar(ref):
self._has_output_scaling |= ref != 1.0
else:
self._has_output_scaling |= np.any(ref != 1.0)
if np.isscalar(ref0):
self._has_output_scaling |= ref0 != 0.0
else:
self._has_output_scaling |= np.any(ref0)
if np.isscalar(res_ref):
self._has_resid_scaling |= res_ref != 1.0
else:
self._has_resid_scaling |= np.any(res_ref != 1.0)
ref = format_as_float_or_array('ref', ref, flatten=True)
ref0 = format_as_float_or_array('ref0', ref0, flatten=True)
res_ref = format_as_float_or_array('res_ref', res_ref, flatten=True)
# ref, ref0, res_ref: taken as is
metadata['ref'] = ref
metadata['ref0'] = ref0
metadata['res_ref'] = res_ref
# var_set: taken as is
metadata['var_set'] = var_set
metadata['distributed'] = self.distributed
# We may not know the pathname yet, so we have to use name for now, instead of abs_name.
if self._static_mode:
var_rel2data_io = self._static_var_rel2data_io
var_rel_names = self._static_var_rel_names
else:
var_rel2data_io = self._var_rel2data_io
var_rel_names = self._var_rel_names
# Disallow dupes
if name in var_rel2data_io:
msg = "Variable name '{}' already exists.".format(name)
raise ValueError(msg)
var_rel2data_io[name] = {
'prom': name,
'rel': name,
'my_idx': len(self._var_rel_names['output']),
'type': 'output',
'metadata': metadata
}
var_rel_names['output'].append(name)
return metadata
def prepare_for_inversion(gdir,
add_debug_var=False,
invert_with_rectangular=True,
invert_all_rectangular=False):
"""Prepares the data needed for the inversion.
Mostly the mass flux and slope angle, the rest (width, height) was already
computed. It is then stored in a list of dicts in order to be faster.
Parameters
----------
gdir : :py:class:`oggm.GlacierDirectory`
the glacier directory to process
"""
# variables
fls = gdir.read_pickle('inversion_flowlines')
towrite = []
for fl in fls:
# Distance between two points
dx = fl.dx * gdir.grid.dx
# Widths
widths = fl.widths * gdir.grid.dx
# Heights
hgt = fl.surface_h
angle = -np.gradient(hgt, dx) # beware the minus sign
# Flux needs to be in [m3 s-1] (*ice* velocity * surface)
# fl.flux is given in kg m-2 yr-1, rho in kg m-3, so this should be it:
rho = cfg.PARAMS['ice_density']
flux = fl.flux * (gdir.grid.dx**2) / cfg.SEC_IN_YEAR / rho
# Clip flux to 0
if np.any(flux < -0.1):
log.warning('(%s) has negative flux somewhere', gdir.rgi_id)
flux = flux.clip(0)
if fl.flows_to is None and gdir.inversion_calving_rate == 0:
if not np.allclose(flux[-1], 0., atol=0.1):
# TODO: this test doesn't seem meaningful here
msg = ('({}) flux at terminus should be zero, but is: '
'{.4f} m3 ice s-1'.format(gdir.rgi_id, flux[-1]))
raise RuntimeError(msg)
flux[-1] = 0.
# Shape
is_rectangular = fl.is_rectangular
if not invert_with_rectangular:
is_rectangular[:] = False
if invert_all_rectangular:
is_rectangular[:] = True
# Optimisation: we need to compute this term of a0 only once
flux_a0 = np.where(is_rectangular, 1, 1.5)
flux_a0 *= flux / widths
# Add to output
cl_dic = dict(dx=dx,
flux_a0=flux_a0,
width=widths,
slope_angle=angle,
is_rectangular=is_rectangular,
is_last=fl.flows_to is None)
if add_debug_var:
cl_dic['flux'] = flux
cl_dic['hgt'] = hgt
towrite.append(cl_dic)
# Write out
gdir.write_pickle(towrite, 'inversion_input')
def signaltools_detrend(data, axis=-1, type='linear', bp=0):
"""
Remove linear trend along axis from data.
Parameters
----------
data : array_like
The input data.
axis : int, optional
The axis along which to detrend the data. By default this is the
last axis (-1).
type : {'linear', 'constant'}, optional
The type of detrending. If ``type == 'linear'`` (default),
the result of a linear least-squares fit to `data` is subtracted
from `data`.
If ``type == 'constant'``, only the mean of `data` is subtracted.
bp : array_like of ints, optional
A sequence of break points. If given, an individual linear fit is
performed for each part of `data` between two break points.
Break points are specified as indices into `data`.
Returns
-------
ret : ndarray
The detrended input data.
"""
if type not in ['linear', 'l', 'constant', 'c']:
raise ValueError("Trend type must be 'linear' or 'constant'.")
data = np.asarray(data)
dtype = data.dtype.char
if dtype not in 'dfDF':
dtype = 'd'
if type in ['constant', 'c']:
#print('Removing mean')
ret = data - np.expand_dims(np.mean(data, axis), axis)
return ret
else:
#print('Removing linear?')
dshape = data.shape
N = dshape[axis]
bp = sort(unique(r_[0, bp, N]))
if np.any(bp > N):
raise ValueError("Breakpoints must be less than length "
"of data along given axis.")
Nreg = len(bp) - 1
# Restructure data so that axis is along first dimension and
# all other dimensions are collapsed into second dimension
rnk = len(dshape)
if axis < 0:
axis = axis + rnk
newdims = r_[axis, 0:axis, axis + 1:rnk]
newdata = reshape(np.transpose(data, tuple(newdims)),
(N, _prod(dshape) // N))
newdata = newdata.copy() # make sure we have a copy
if newdata.dtype.char not in 'dfDF':
newdata = newdata.astype(dtype)
# Find leastsq fit and remove it for each piece
for m in range(Nreg):
Npts = bp[m + 1] - bp[m]
A = ones((Npts, 2), dtype)
A[:, 0] = cast[dtype](np.arange(1, Npts + 1) * 1.0 / Npts)
sl = slice(bp[m], bp[m + 1])
coef, resids, rank, s = np.linalg.lstsq(A, newdata[sl])
newdata[sl] = newdata[sl] - dot(A, coef)
# Put data back in original shape.
tdshape = take(dshape, newdims, 0)
ret = np.reshape(newdata, tuple(tdshape))
vals = list(range(1, rnk))
olddims = vals[:axis] + [0] + vals[axis:]
ret = np.transpose(ret, tuple(olddims))
return ret
def partition(a):
"""
given a set, determine if you can partition into two sets s.t. the sum
is the same
ok so if you sum these numbers and they are odd then it's impossible
so first you compute the sum and check that
so now you have an even sum what do you do?
can you view this as a graph?
yeah of course but what's the graph
dag?
not sure
so for each number you can either put it in set 1 or set 2
what's the subproblem?
at each point in the list know whether you can separte up tot hat point
evenly?
seems like it takes the same form as the longest nondecreasing
but why?
find a subset equal to sum/2 b/c the other subset must therefore equal sum/2
as well
ok h = sum(s) / 2
for each number decide whether to add it to the set or not
ok
2d array
size sum/2 x n+1
where the rows correspond to sums
where the cols correspond to whether or not a subset starting at this location
and moving backward can sum up to the value
how would you fill this?
[2,1,1]
originally
[0,0,0]
[0,0,0]
[0,1,1]
[0,0,0]
[0,0,0]
[0,0,0]
how to fill (2,1)? 0 based indexing
- not sure
for each row you go through the entire sequence
I think it works like this
for each possible sum value
you see if you could add up the elements to that sum
the way you do this efficiently is you if the current sum value (call that i)
minus the value you're going to add (v[j-1]) could add up to whatever sum
results
then it should work
so
looking at that example
[2,1,1]
0 [1,1,1,1] # all ones here indicate that of course you can add to 0
1 [0,0,0,0]
2 [0,0,0,0]
deciding the first value
value is 2
is 2 less than the current sum? no then no
0 [1,1,1,1]
1 [0,0,0,0]
2 [0,0,0,0]
next look at 1
first of all, if you could have added to the current sum by the
prior column
then you can still add to that value by not considering this value
so initialize it to the value in the column before which here is 0
the other condition whereby this cell might be true is that
where the current sum minus the current value is achievable without using
the current value
current sum = 1
current value = 1
table[0,]
0 [1,1,1,1]
1 [0,0,0,0]
2 [0,0,0,0]
ok so how does this work?
it relies on the fact that the sum is integer only
and you basically find out whether you can sum to any value that is
less than half of the total
it does this by creating a table where the rows are the possible sums
i.e., all values < sum/2
and the rows correspond to the elements plus a row without anything
so example
[2,1,1]
sum/2 = 2
table
{} {2} {2,1} {2,11} # so the cols reflect the subsets up to that point
# and whether that subset can add to that sum
0 [1,1,1,1] # b/c anything can add to 0 if you don't take it
1 [0,0,0,0]
2 [0,0,0,0]
# initialize each new cell to the cell left of it
# because if the previous set could add to the value
# then the new set can as well by not using the new value
# to fill in (1, {2,1})
# look at the row minus the element (1 - 1) = 0 // this row
# then to know whether you could sum to this value at this column
# you ask without this column could you sum to this sum minus this column?
the answer to this question is in part[i - arr[j-1]][j-1]
this indexing is tricky
j-1 indexes into the array where the jth element is
whereas it indexes into the container where the sum without this value is
"""
a, s = np.array(a), np.sum(a)
if s % 2 == 1:
return False, []
s, n = int(s/2), len(a)
t = np.zeros((s+1, n+1))
t[0,:] = 1
for i in range(1, s+1):
# j indexes into t for the j-1 element of a
for j in range(1, n+1):
# if you can sum to the value without this col, then you can with it
t[i,j] = t[i,j-1]
if s >= a[j-1]:
# again if t[i,j-1] true then stay true
# become true if the current sum minus the current col value
# can be produced using value up to but not including the current
# value
t[i,j] = t[i,j] or t[i-a[j-1]][j-1]
print(t)
# how do you get the partitioning?
# move left until final true value
# subtract that value
# repeat
# gives the indices of the values contributing to the sum
i, j, idxs = s, n, []
while True:
if i == 0:
break
while True:
if t[i,j-1]:
j -=1
else:
break
# a[j-1] is part of group
idxs.append(j-1)
i -= a[j-1]
# if you want to know whether you can do the partition, just return
# whether anything in the last row is true
return np.any(t[-1,:]), idxs
def _inBoard(self, pos):
return not np.any([pos < [0, 0], pos >= self._map.shape])
[-LIGHT_X_INSIDE, -0.2, BACK_Z], # 26
[-LIGHT_X_INSIDE + 0.1, -0.3, BACK_Z], # 27
[-X_OUTSIDE + 0.1, BOTTOM_LINE, BACK_Z]] + \
[[np.nan, np.nan, np.nan]] * 30 + \
[[-P_X, P_Y_TOP, FRONT_Z]] + \
[[np.nan, np.nan, np.nan]] + \
[[-P_X, P_Y_BOTTOM, FRONT_Z], # 61
[-P_X, P_Y_TOP, BACK_Z]] + \
[[np.nan, np.nan, np.nan]] * 2 + \
[[-P_X, P_Y_BOTTOM, BACK_Z]]) # 65
CAR_POSE_66 = CAR_POSE_HALF
for key in HFLIP_ids:
CAR_POSE_66[HFLIP_ids[key], :] = CAR_POSE_HALF[key, :]
CAR_POSE_66[HFLIP_ids[key], 0] = -CAR_POSE_HALF[key, 0]
assert not np.any(CAR_POSE_66 == np.nan)
training_weights_local_centrality = [
0.890968488270775, 0.716506138617812, 1.05674590410869, 0.764774195768455,
0.637682585483328, 0.686680807728366, 0.955422595797394, 0.936714585642375,
1.34823795445326, 1.38308992581967, 1.32689945125819, 1.38838655605483,
1.18980184904613, 1.02584355494795, 0.90969156732068, 1.24732068576104,
1.11338768064342, 0.933815217550391, 0.852297518872114, 1.04167641424727,
1.01668968075247, 1.34625964088011, 0.911796331039028, 0.866206536337413,
1.55957820407853, 0.730844382675724, 0.651138644197359, 0.758018559633786,
1.31842501396691, 1.32186116654782, 0.744347016851606, 0.636390683664723,
0.715244950821949, 1.63122349407032, 0.849835699185461, 0.910488007220499,
1.44244151650561, 1.14150437331681, 1.19808610191343, 0.960186788642886,
1.05023623286937, 1.19761709710598, 1.3872216313401, 1.01256700741214,
1.1167909667759, 1.27893496336199, 1.54475684725655, 1.40343733870633,
1.45552060866114, 1.47264222155031, 0.970060423999993, 0.944450314768933,
def __contains__(self, key):
try:
res = self.get_loc(key)
return np.isscalar(res) or type(res) == slice or np.any(res)
except (KeyError, TypeError, ValueError):
return False
def _eq(self, pos1, pos2):
return not np.any(pos1 != pos2)
def calc_slope_vars(rn_sect, gain_sect, gdq_sect, group_time, max_seg):
"""
Calculate the segment-specific variance arrays for the given
integration.
Parameters
----------
rn_sect : ndarray
read noise values for all pixels in data section, 2-D float
gain_sect : ndarray
gain values for all pixels in data section, 2-D float
gdq_sect : ndarray
data quality flags for pixels in section, 3-D int
group_time : float
Time increment between groups, in seconds.
max_seg : int
maximum number of segments fit
Returns
-------
den_r3 : ndarray
for a given integration, the reciprocal of the denominator of the
segment-specific variance of the segment's slope due to read noise, 3-D float
den_p3 : ndarray
for a given integration, the reciprocal of the denominator of the
segment-specific variance of the segment's slope due to Poisson noise, 3-D float
num_r3 : ndarray
numerator of the segment-specific variance of the segment's slope
due to read noise, 3-D float
segs_beg_3 : ndarray
lengths of segments for all pixels in the given data section and
integration, 3-D int
"""
(nreads, asize2, asize1) = gdq_sect.shape
npix = asize1 * asize2
imshape = (asize2, asize1)
# Create integration-specific sections of input arrays for determination
# of the variances.
gdq_2d = gdq_sect[:, :, :].reshape((nreads, npix))
gain_1d = gain_sect.reshape(npix)
gdq_2d_nan = gdq_2d.copy() # group dq with SATS will be replaced by nans
gdq_2d_nan = gdq_2d_nan.astype(np.float32)
wh_sat = np.where(np.bitwise_and(gdq_2d, constants.dqflags["SATURATED"]))
if len(wh_sat[0]) > 0:
gdq_2d_nan[wh_sat] = np.nan # set all SAT groups to nan
del wh_sat
# Get lengths of semiramps for all pix [number_of_semiramps, number_of_pix]
segs = np.zeros_like(gdq_2d)
# Counter of semiramp for each pixel
sr_index = np.zeros(npix, dtype=np.uint8)
pix_not_done = np.ones(npix, dtype=bool) # initialize to True
i_read = 0
# Loop over reads for all pixels to get segments (segments per pixel)
while (i_read < nreads and np.any(pix_not_done)):
gdq_1d = gdq_2d_nan[i_read, :]
wh_good = np.where(gdq_1d == 0) # good groups
# if this group is good, increment those pixels' segments' lengths
if len(wh_good[0]) > 0:
segs[sr_index[wh_good], wh_good] += 1
del wh_good
# Locate any CRs that appear before the first SAT group...
wh_cr = np.where(gdq_2d_nan[i_read, :].astype(np.int32)
& constants.dqflags["JUMP_DET"] > 0)
# ... but not on final read:
if (len(wh_cr[0]) > 0 and (i_read < nreads - 1)):
sr_index[wh_cr[0]] += 1
segs[sr_index[wh_cr], wh_cr] += 1
del wh_cr
# If current group is a NaN, this pixel is done (pix_not_done is False)
wh_nan = np.where(np.isnan(gdq_2d_nan[i_read, :]))
if len(wh_nan[0]) > 0:
pix_not_done[wh_nan[0]] = False
del wh_nan
i_read += 1
segs = segs.astype(np.uint8)
segs_beg = segs[:max_seg, :] # the leading nonzero lengths
# Create reshaped version [ segs, y, x ] to simplify computation
segs_beg_3 = segs_beg.reshape(max_seg, imshape[0], imshape[1])
segs_beg_3 = remove_bad_singles(segs_beg_3)
# Create a version 1 less for later calculations for the variance due to
# Poisson, with a floor=1 to handle single-group segments
wh_pos_3 = np.where(segs_beg_3 > 1)
segs_beg_3_m1 = segs_beg_3.copy()
segs_beg_3_m1[wh_pos_3] -= 1
segs_beg_3_m1[segs_beg_3_m1 < 1] = 1
# For a segment, the variance due to Poisson noise
# = slope/(tgroup * gain * (ngroups-1)),
# where slope is the estimated median slope, tgroup is the group time,
# and ngroups is the number of groups in the segment.
# Here the denominator of this quantity will be computed, which will be
# later multiplied by the estimated median slope.
# Suppress, then re-enable, harmless arithmetic warnings, as NaN will be
# checked for and handled later
warnings.filterwarnings("ignore", ".*invalid value.*", RuntimeWarning)
warnings.filterwarnings("ignore", ".*divide by zero.*", RuntimeWarning)
den_p3 = 1. / (group_time * gain_1d.reshape(imshape) * segs_beg_3_m1)
warnings.resetwarnings()
# For a segment, the variance due to readnoise noise
# = 12 * readnoise**2 /(ngroups_seg**3. - ngroups_seg)/( tgroup **2.)
num_r3 = 12. * (rn_sect / group_time)**2. # always >0
# Reshape for every group, every pixel in section
num_r3 = np.dstack([num_r3] * max_seg)
num_r3 = np.transpose(num_r3, (2, 0, 1))
# Denominator den_r3 = 1./(segs_beg_3 **3.-segs_beg_3). The minimum number
# of allowed groups is 2, which will apply if there is actually only 1
# group; in this case den_r3 = 1/6. This covers the case in which there is
# only one good group at the beginning of the integration, so it will be
# be compared to the plane of (near) zeros resulting from the reset. For
# longer segments, this value is overwritten below.
den_r3 = num_r3.copy() * 0. + 1. / 6
wh_seg_pos = np.where(segs_beg_3 > 1)
# Suppress, then, re-enable harmless arithmetic warnings, as NaN will be
# checked for and handled later
warnings.filterwarnings("ignore", ".*invalid value.*", RuntimeWarning)
warnings.filterwarnings("ignore", ".*divide by zero.*", RuntimeWarning)
den_r3[wh_seg_pos] = 1. / (
segs_beg_3[wh_seg_pos]**3. - segs_beg_3[wh_seg_pos]
) # overwrite where segs>1
warnings.resetwarnings()
return (den_r3, den_p3, num_r3, segs_beg_3)
"Selección de fechas en Noviembre, Diciembre, Enero, Febrero, Marzo, Abril"
pos_2016_04_30 = np.where(fechas == Timestamp('2016-04-30 18:00:00'))[0][0]
FCH = fechas[3: pos_2016_04_30 +1 : 4]
DT = []
leap_years = np.array([x for x in set(FCH[FCH.is_leap_year].year)]) # Años bisiestos
normal_years = np.array([x for x in set(FCH[~FCH.is_leap_year].year)]) # Años normales
years = np.array([x for x in set(FCH.year)])
for i in years[:-1]:
pos = np.where(FCH == pd.Timestamp(str(i)+'-11-01 18:00:00'))[0][0]
if np.any(normal_years == i+1) == True:
DT.append(FCH[pos:pos+181])
else:
DT.append(FCH[pos:pos+182])
FECHAS = pd.DatetimeIndex(np.concatenate(DT))
"Lectura de Estados"
rf = open('/home/yordan/YORDAN/UNAL/TESIS_MAESTRIA/25_expo_2018/States_'+ch+'_NovAbr_anom_925.csv', 'r')
reader = csv.reader(rf)
states = [row for row in reader][1:]
rf.close()
print "La extracción de los estados en la exposición 26 para hacer composites de presión, se hace sin los últimos 181 datos, ya que los datos que se tienen de presión sólo llegan hasta el 2016-01-31"
states6 = np.array([int(x[5]) for x in states])[:-181]