def test_rate_arrays():
P = PoissonGroup(2, np.array([0, 1./defaultclock.dt])*Hz)
spikes = SpikeMonitor(P)
net = Network(P, spikes)
net.run(2*defaultclock.dt)
assert_equal(spikes.count, np.array([0, 2]))
def test_multinomial_elementwise_vector(self):
'''Test creating multinomial variables (r=1).'''
(m, n) = (20, 5)
p = statutil.random_row_stochastic((m, n))
x = statutil.multinomial_elementwise(p)
assert_equal(x, [3, 1, 3, 4, 0, 2, 2, 3, 1, 4, 2, 3, 2, 1, 4, 2, 4, 3, 4, 3],
'Wrong random multinomial generated')
def test_file(file_name, d=None):
n = 50
for d in [d] if d else xrange(1, n + 5):
fast = sum_from_left(read_digit_columns(open(file_name, 'rb'), n), d)
exact = int(str(sum(int(x.rstrip('\n')) for x in open(file_name, 'rb')))[0:d])
# print d, fast, exact
assert_equal(fast, exact, 'Approximate sum is wrong, file %s, d %d' % (file_name, d))
def test_spike_monitor():
language_before = brian_prefs.codegen.target
for language in languages:
brian_prefs.codegen.target = language
defaultclock.t = 0*second
G = NeuronGroup(2, '''dv/dt = rate : 1
rate: Hz''', threshold='v>1', reset='v=0')
# We don't use 100 and 1000Hz, because then the membrane potential would
# be exactly at 1 after 10 resp. 100 timesteps. Due to floating point
# issues this will not be exact,
G.rate = [101, 1001] * Hz
mon = SpikeMonitor(G)
net = Network(G, mon)
net.run(10*ms)
assert_allclose(mon.t[mon.i == 0], [9.9]*ms)
assert_allclose(mon.t[mon.i == 1], np.arange(10)*ms + 0.9*ms)
assert_allclose(mon.t_[mon.i == 0], np.array([9.9*float(ms)]))
assert_allclose(mon.t_[mon.i == 1], (np.arange(10) + 0.9)*float(ms))
assert_equal(mon.count, np.array([1, 10]))
i, t = mon.it
i_, t_ = mon.it_
assert_equal(i, mon.i)
assert_equal(i, i_)
assert_equal(t, mon.t)
assert_equal(t_, mon.t_)
brian_prefs.codegen.target = language_before
def test_changed_dt_spikes_in_queue():
for codeobj_class in codeobj_classes:
defaultclock.dt = .5*ms
G1 = NeuronGroup(1, 'v:1', threshold='v>1', reset='v=0',
codeobj_class=codeobj_class)
G1.v = 1.1
G2 = NeuronGroup(10, 'v:1', threshold='v>1', reset='v=0',
codeobj_class=codeobj_class)
S = Synapses(G1, G2, pre='v+=1.1', codeobj_class=codeobj_class)
S.connect(True)
S.delay = 'j*ms'
mon = SpikeMonitor(G2)
net = Network(G1, G2, S, mon)
net.run(5*ms)
defaultclock.dt = 1*ms
net.run(3*ms)
defaultclock.dt = 0.1*ms
net.run(2*ms)
# Spikes should have delays of 0, 1, 2, ... ms and always
# trigger a spike one dt later
expected = [0.5, 1.5, 2.5, 3.5, 4.5, # dt=0.5ms
6, 7, 8, #dt = 1ms
8.1, 9.1 #dt=0.1ms
] * ms
assert_equal(mon.t, expected)
def test_LinearConstraint():
from numpy.testing.utils import assert_equal
lc = LinearConstraint(["foo", "bar"], [1, 1])
assert lc.variable_names == ["foo", "bar"]
assert_equal(lc.coefs, [[1, 1]])
assert_equal(lc.constants, [[0]])
lc = LinearConstraint(["foo", "bar"], [[1, 1], [2, 3]], [10, 20])
assert_equal(lc.coefs, [[1, 1], [2, 3]])
assert_equal(lc.constants, [[10], [20]])
assert lc.coefs.dtype == np.dtype(float)
assert lc.constants.dtype == np.dtype(float)
# statsmodels wants to be able to create degenerate constraints like this,
# see:
# https://github.com/pydata/patsy/issues/89
# We used to forbid it, but I guess it's harmless, so why not.
lc = LinearConstraint(["a"], [[0]])
assert_equal(lc.coefs, [[0]])
from nose.tools import assert_raises
assert_raises(ValueError, LinearConstraint, ["a"], [[1, 2]])
assert_raises(ValueError, LinearConstraint, ["a"], [[[1]]])
assert_raises(ValueError, LinearConstraint, ["a"], [[1, 2]], [3, 4])
assert_raises(ValueError, LinearConstraint, ["a", "b"], [[1, 2]], [3, 4])
assert_raises(ValueError, LinearConstraint, ["a"], [[1]], [[]])
assert_raises(ValueError, LinearConstraint, ["a", "b"], [])
assert_raises(ValueError, LinearConstraint, ["a", "b"],
np.zeros((0, 2)))
assert_no_pickling(lc)
def test_timedarray_with_units():
ta = TimedArray(np.arange(10)*amp, dt=0.1*ms)
G = NeuronGroup(1, 'value = ta(t) + 2*nA: amp', dt=0.1*ms)
mon = StateMonitor(G, 'value', record=True, dt=0.1*ms)
net = Network(G, mon)
net.run(1.1*ms)
assert_equal(mon[0].value, np.clip(np.arange(len(mon[0].t)), 0, 9)*amp + 2*nA)
def test_spikegenerator_rounding():
# all spikes should fall into the first time bin
indices = np.arange(100)
times = np.linspace(0, 0.1, 100, endpoint=False)*ms
SG = SpikeGeneratorGroup(100, indices, times, dt=0.1*ms)
mon = SpikeMonitor(SG)
net = Network(SG, mon)
net.run(0.1*ms)
assert_equal(mon.count, np.ones(100))
# all spikes should fall in separate bins
dt = 0.1*ms
indices = np.zeros(10000)
times = np.arange(10000)*dt
SG = SpikeGeneratorGroup(1, indices, times, dt=dt)
target = NeuronGroup(1, 'count : 1',
threshold='True', reset='count=0') # set count to zero at every time step
syn = Synapses(SG, target, on_pre='count+=1')
syn.connect()
mon = StateMonitor(target, 'count', record=0, when='end')
net = Network(SG, target, syn, mon)
# change the schedule so that resets are processed before synapses
net.schedule = ['start', 'groups', 'thresholds', 'resets', 'synapses', 'end']
net.run(10000*dt)
assert_equal(mon[0].count, np.ones(10000))
def test_no_reference_1():
'''
Using subgroups without keeping an explicit reference. Basic access.
'''
G = NeuronGroup(10, 'v:1')
G.v = np.arange(10)
assert_equal(G[:5].v[:], G.v[:5])
def testDirectedIntersectionFromInside(self):
# line origin is inside sphere/ellipsoid
for ellipsoidLineIntersection, ellipsoidLineIntersects in zip(ellipsoidLineIntersectionFns, ellipsoidLineIntersectsFns):
r = 2
origin = [1,0,0]
direction = [[1,0,0]]
intersection = [[2,0,0]]
res = sphereLineIntersection(r, origin, direction, directed=False)
assert_array_equal(res, intersection)
res = sphereLineIntersection(r, origin, direction, directed=True)
assert_array_equal(res, intersection)
res = ellipsoidLineIntersection(r, r, origin, direction, directed=False)
assert_array_equal(res, intersection)
intersects = ellipsoidLineIntersects(r, r, origin, direction, directed=False)
assert_equal(intersects, [True])
res = ellipsoidLineIntersection(r, r, origin, direction, directed=True)
assert_array_equal(res, intersection)
intersects = ellipsoidLineIntersects(r, r, origin, direction, directed=True)
assert_equal(intersects, [True])
intersection2 = [[-2,0,0]]
direction2 = [[-1,0,0]]
res = sphereLineIntersection(r, origin, direction2, directed=False)
assert_array_equal(res, intersection)
res = sphereLineIntersection(r, origin, direction2, directed=True)
assert_array_equal(res, intersection2)
def test_decompose_pentagon(self):
# Make sure decompose returns correct thing for a pentagon.
result = decompose_to_triangles(self.pentagon)
benchmark = array([[[0, 0], [2, 0], [2, 1]],
[[0, 0], [2, 1], [1, 2]],
[[0, 0], [1, 2], [0, 2]]], dtype='float')
assert_equal(result, benchmark)
def test_polygon1():
vert = [(2, 1), (3, 5), (6, 6), (3, 8), (0, 4), (2, 1)]
pol = region.Polygon('1', vert)
mask = np.zeros((9, 9), dtype=np.int)
mask = pol.scan(mask)
pol1 = polygon1()
utils.assert_equal(mask, pol1)
def testSphereLineIntersection(self):
sphereRadius = 2
lineOrigin = [0,3,0]
lineDirection = [0,-1,0]
point = sphereLineIntersection(sphereRadius, lineOrigin, lineDirection)
assert_equal(point, [0,2,0])
def testSphereLineIntersectionArray(self):
sphereRadius = 2
lineOrigin = [0,3,0]
lineDirection = unitVectors([[0,-1,0],[-1,-1,0]])
points = sphereLineIntersection(sphereRadius, lineOrigin, lineDirection)
assert_equal(points, [[0,2,0],[np.nan,np.nan,np.nan]])
def test_scalar_parameter_access():
for codeobj_class in codeobj_classes:
G = NeuronGroup(10, '''dv/dt = freq : 1
freq : Hz (shared)
number : 1 (shared)
array : 1''',
codeobj_class=codeobj_class)
# Try setting a scalar variable
G.freq = 100*Hz
assert_equal(G.freq[:], 100*Hz)
G.freq[:] = 200*Hz
assert_equal(G.freq[:], 200*Hz)
G.freq = 'freq - 50*Hz + number*Hz'
assert_equal(G.freq[:], 150*Hz)
G.freq[:] = '50*Hz'
assert_equal(G.freq[:], 50*Hz)
# Check the second method of accessing that works
assert_equal(np.asanyarray(G.freq), 50*Hz)
# Check error messages
assert_raises(IndexError, lambda: G.freq[0])
assert_raises(IndexError, lambda: G.freq[1])
assert_raises(IndexError, lambda: G.freq[0:1])
assert_raises(IndexError, lambda: G.freq['i>5'])
assert_raises(ValueError, lambda: G.freq.set_item(slice(None), [0, 1]*Hz))
assert_raises(IndexError, lambda: G.freq.set_item(0, 100*Hz))
assert_raises(IndexError, lambda: G.freq.set_item(1, 100*Hz))
assert_raises(IndexError, lambda: G.freq.set_item('i>5', 100*Hz))
def test_par_slicing(self):
"""
Test assigning to a parameter slice
"""
p1 = models.Polynomial1D(3, n_models=3)
p1.c0[:2] = [10, 10]
utils.assert_equal(p1.parameters, [10.0, 10.0, 0.0, 0, 0, 0, 0, 0, 0, 0, 0, 0])
def test_synapse_creation_generator_complex_ranges():
G1 = NeuronGroup(10, 'v:1', threshold='False')
G2 = NeuronGroup(20, 'v:1', threshold='False')
G1.v = 'i'
G2.v = '10 + i'
SG1 = G1[:5]
SG2 = G2[10:]
S = Synapses(SG1, SG2, 'w:1', on_pre='v+=w')
S.connect(j='i+k for k in range(N_post-i)') # Connect to all j>i
# connect based on pre-/postsynaptic state variables
S2 = Synapses(SG1, SG2, 'w:1', on_pre='v+=w')
S2.connect(j='k for k in range(N_post * int(v_pre > 2))')
# connect based on pre-/postsynaptic state variables
S3 = Synapses(SG2, SG1, 'w:1', on_pre='v+=w')
S3.connect(j='k for k in range(N_post * int(v_pre > 22))')
run(0*ms) # for standalone
for syn_source in xrange(5):
# Internally, the "real" neuron indices should be used
assert_equal(S._synaptic_post[syn_source, :],
10 + syn_source + np.arange(10 - syn_source))
# For the user, the subgroup-relative indices should be presented
assert_equal(S.j[syn_source, :], syn_source + np.arange(10-syn_source))
assert len(S2) == 2 * len(SG2), str(len(S2))
assert all(S2.v_pre[:] > 2)
assert len(S3) == 7 * len(SG1), str(len(S3))
assert all(S3.v_pre[:] > 22)
def test_compute_ts_map(tmpdir):
"""Minimal test of compute_ts_map"""
data = load_poisson_stats_image(extra_info=True)
kernel = Gaussian2DKernel(2.5)
data['exposure'] = np.ones(data['counts'].shape) * 1E12
for _, func in zip(['counts', 'background', 'exposure'], [np.nansum, np.nansum, np.mean]):
data[_] = downsample_2N(data[_], 2, func)
result = compute_ts_map(data['counts'], data['background'], data['exposure'],
kernel)
for name, order in zip(['ts', 'amplitude', 'niter'], [2, 5, 0]):
result[name] = np.nan_to_num(result[name])
result[name] = upsample_2N(result[name], 2, order=order)
assert_allclose(1705.840212274973, result.ts[99, 99], rtol=1e-3)
assert_allclose([[99], [99]], np.where(result.ts == result.ts.max()))
assert_allclose(6, result.niter[99, 99])
assert_allclose(1.0227934338735763e-09, result.amplitude[99, 99], rtol=1e-3)
# test write method
filename = str(tmpdir.join('ts_test.fits'))
result.write(filename, header=data['header'])
read_result = TSMapResult.read(filename)
for _ in ['ts', 'sqrt_ts', 'amplitude', 'niter']:
assert result[_].dtype == read_result[_].dtype
assert_equal(result[_], read_result[_])
def read_pedigree_from_test_file(file_name, genotyped_id_file=None):
'''Load a pedigree from a PLINK TFAM file.'''
data = np.genfromtxt(file_name, np.dtype(int))
p = io_pedigree.read(file_name, genotyped_id_file=genotyped_id_file)
assert_equal(p._graph.number_of_nodes(), data.shape[0], 'Incorrect number of nodes')
assert nx.is_directed_acyclic_graph(p._graph), 'Pedigree is not a DAG'
return p
def test_timedarray_no_upsampling():
# Test a TimedArray where no upsampling is necessary because the monitor's
# dt is bigger than the TimedArray's
ta = TimedArray(np.arange(10), dt=0.01*ms)
G = NeuronGroup(1, 'value = ta(t): 1', dt=0.1*ms)
mon = StateMonitor(G, 'value', record=True, dt=1*ms)
run(2.1*ms)
assert_equal(mon[0].value, [0, 9, 9])
def test_state_variables_group_as_index():
G = NeuronGroup(10, 'v : 1')
SG = G[4:9]
G.v[SG] = 1
assert_equal(G.v[:], np.array([0, 0, 0, 0, 1, 1, 1, 1, 1, 0]))
G.v = 1
G.v[SG] = '2*v'
assert_equal(G.v[:], np.array([1, 1, 1, 1, 2, 2, 2, 2, 2, 1]))
def __test_segment_range(self, n, step, c):
'''Test getting the start and end of equidistant intervals in an item collection.'''
items = np.arange(0, n, c)
segments = iu.segmentrange(items, step)
assert_equal(segments[:, 0], range(0, n, c * step), 'Wrong start array')
assert_equal(segments[:, 1],
range(c * (step - 1), n, c * step) + ([n - c] if np.mod(n, step) else []),
'Wrong end array')
def test_poly1d_multiple_sets(self):
p1 = models.Polynomial1D(3, n_models=3)
utils.assert_equal(p1.parameters, [0.0, 0.0, 0.0, 0, 0, 0,
0, 0, 0, 0, 0, 0])
utils.assert_array_equal(p1.c0, [0, 0, 0])
p1.c0 = [10, 10, 10]
utils.assert_equal(p1.parameters, [10.0, 10.0, 10.0, 0, 0,
0, 0, 0, 0, 0, 0, 0])
def test_propagation():
# Using a PoissonGroup as a source for Synapses should work as expected
P = PoissonGroup(2, np.array([0, 1.0 / defaultclock.dt]) * Hz)
G = NeuronGroup(2, "v:1")
S = Synapses(P, G, pre="v+=1", connect="i==j")
run(2 * defaultclock.dt)
assert_equal(G.v[:], np.array([0.0, 2.0]))
def test_par_slicing(self):
"""
Test assigning to a parameter slice
"""
p1 = models.Poly1DModel(3, param_dim=3)
p1.c0[:2] = [10, 10]
utils.assert_equal(p1.parameters, [10.0, 10.0, 0.0, 0, 0,
0, 0, 0, 0, 0, 0, 0])
def test_timedarray_with_units():
ta = TimedArray(np.arange(10)*amp, defaultclock.dt)
for codeobj_class in codeobj_classes:
G = NeuronGroup(1, 'value = ta(t) + 2*nA: amp', codeobj_class=codeobj_class)
mon = StateMonitor(G, 'value', record=True)
net = Network(G, mon)
net.run(11*ms)
assert_equal(mon[0].value, np.clip(np.arange(len(mon[0].t)), 0, 9)*amp + 2*nA)
def test_propagation():
# Using a PoissonGroup as a source for Synapses should work as expected
P = PoissonGroup(2, np.array([0, 1./defaultclock.dt])*Hz)
G = NeuronGroup(2, 'v:1')
S = Synapses(P, G, pre='v+=1', connect='i==j')
run(2*defaultclock.dt)
assert_equal(G.v[:], np.array([0., 2.]))
def test_subexpression():
G = NeuronGroup(10, '''dv/dt = freq : 1
freq : Hz
array : 1
expr = 2*freq + array*Hz : Hz''')
G.freq = '10*i*Hz'
G.array = 5
assert_equal(G.expr[:], 2*10*np.arange(10)*Hz + 5*Hz)
def test_non_linear_NYset(self):
"""
This case covers:
N param sets , 1 set 1D x --> N 1D y data
"""
g1 = models.Gaussian1DModel([10, 10], [3, 3], [.2, .2])
y1 = g1(self.x1)
utils.assert_equal((y1[:, 0] - y1[:, 1]).nonzero(), (np.array([]),))
def test_alpha(self):
""" This test checks if RidgeCV finds the optimal `alpha`.
"""
self.var.fit(self.x)
# Currently we simply *know* empirically that from the three
# candidate alphas 100 is closest to the optimum.
# TODO: programmatically derive the optimum from the data
assert_equal(self.var.fitting_model.alpha_, 100)
def test_refractoriness_variables():
# Try a string evaluating to a quantity an an explicit boolean
# condition -- all should do the same thing
for ref_time in [
'5*ms', '(t-lastspike) <= 5*ms', 'time_since_spike <= 5*ms',
'ref_subexpression', '(t-lastspike) <= ref', 'ref',
'ref_no_unit*ms'
]:
G = NeuronGroup(1,
'''
dv/dt = 100*Hz : 1 (unless refractory)
dw/dt = 100*Hz : 1
ref : second
ref_no_unit : 1
time_since_spike = t - lastspike : second
ref_subexpression = (t - lastspike) <= ref : boolean
''',
threshold='v>1',
reset='v=0;w=0',
refractory=ref_time)
G.ref = 5 * ms
G.ref_no_unit = 5
# It should take 10ms to reach the threshold, then v should stay at 0
# for 5ms, while w continues to increase
mon = StateMonitor(G, ['v', 'w'], record=True)
net = Network(G, mon)
net.run(20 * ms)
try:
# No difference before the spike
assert_equal(mon[0].v[mon.t < 10 * ms], mon[0].w[mon.t < 10 * ms])
# v is not updated during refractoriness
in_refractoriness = mon[0].v[(mon.t >= 10 * ms)
& (mon.t < 15 * ms)]
assert_equal(in_refractoriness, np.zeros_like(in_refractoriness))
# w should evolve as before
assert_equal(mon[0].w[mon.t < 5 * ms],
mon[0].w[(mon.t >= 10 * ms) & (mon.t < 15 * ms)])
assert np.all(mon[0].w[(mon.t >= 10 * ms) & (mon.t < 15 * ms)] > 0)
# After refractoriness, v should increase again
assert np.all(mon[0].v[(mon.t >= 15 * ms) & (mon.t < 20 * ms)] > 0)
except AssertionError as ex:
raise AssertionError(
'Assertion failed when using %r as refractory argument:\n%s' %
(ref_time, ex))
def test_get_features_gamma():
model_spikes = [
[np.array([1, 5, 8]), np.array([2, 3, 8, 9])], # Correct rate
[np.array([1, 5]), np.array([0, 2, 3, 8, 9])]
] # Wrong rate
data_spikes = [np.array([0, 5, 9]), np.array([1, 3, 5, 6])]
gf = GammaFactor(delta=0.5 * ms, time=10 * ms)
features = gf.get_features(model_spikes, data_spikes, 0.1 * ms)
assert_equal(np.shape(features), (2, 2))
assert (np.all(np.array(features) > -1))
normed_gf = GammaFactor(delta=0.5 * ms, time=10 * ms, normalization=1 / 2.)
normed_features = normed_gf.get_features(model_spikes, data_spikes,
0.1 * ms)
assert_equal(normed_features, 2 * features)
features = gf.get_features([data_spikes] * 3, data_spikes, 0.1 * ms)
assert_equal(np.shape(features), (3, 2))
assert_almost_equal(features, np.zeros((3, 2)))
def test_poissoninput():
# Test extreme cases and do a very basic test of an intermediate case, we
# don't want tests to be stochastic
G = NeuronGroup(
10, '''x : volt
y : volt
y2 : volt
z : volt
z2 : volt
w : 1''')
G.w = 0.5
never_update = PoissonInput(G, 'x', 100, 0 * Hz, weight=1 * volt)
always_update = PoissonInput(G,
'y',
50,
1 / defaultclock.dt,
weight=2 * volt)
always_update2 = PoissonInput(G,
'y2',
50,
1 / defaultclock.dt,
weight='1*volt + 1*volt')
sometimes_update = PoissonInput(G, 'z', 10000, 50 * Hz, weight=0.5 * volt)
sometimes_update2 = PoissonInput(G, 'z2', 10000, 50 * Hz, weight='w*volt')
mon = StateMonitor(G, ['x', 'y', 'y2', 'z', 'z2'], record=True, when='end')
run(1 * ms)
assert_equal(0, mon.x[:])
assert_equal(
np.tile((1 + np.arange(mon.y[:].shape[1])) * 50 * 2 * volt, (10, 1)),
mon.y[:])
assert_equal(
np.tile((1 + np.arange(mon.y[:].shape[1])) * 50 * 2 * volt, (10, 1)),
mon.y2[:])
assert all(np.var(mon.z[:], axis=1) > 0) # variability over time
assert all(np.var(mon.z[:], axis=0) > 0) # variability over neurons
assert all(np.var(mon.z2[:], axis=1) > 0) # variability over time
assert all(np.var(mon.z2[:], axis=0) > 0) # variability over neurons
def test_refractoriness_basic():
G = NeuronGroup(1,
'''
dv/dt = 100*Hz : 1 (unless refractory)
dw/dt = 100*Hz : 1
''',
threshold='v>1',
reset='v=0;w=0',
refractory=5 * ms)
# It should take 10ms to reach the threshold, then v should stay at 0
# for 5ms, while w continues to increase
mon = StateMonitor(G, ['v', 'w'], record=True, when='end')
run(20 * ms)
# No difference before the spike
assert_equal(mon[0].v[mon.t < 10 * ms], mon[0].w[mon.t < 10 * ms])
# v is not updated during refractoriness
in_refractoriness = mon[0].v[(mon.t >= 10 * ms) & (mon.t < 15 * ms)]
assert_equal(in_refractoriness, np.zeros_like(in_refractoriness))
# w should evolve as before
assert_equal(mon[0].w[mon.t < 5 * ms],
mon[0].w[(mon.t >= 10 * ms) & (mon.t < 15 * ms)])
assert np.all(mon[0].w[(mon.t >= 10 * ms) & (mon.t < 15 * ms)] > 0)
# After refractoriness, v should increase again
assert np.all(mon[0].v[(mon.t >= 15 * ms) & (mon.t < 20 * ms)] > 0)
def test_unbalanced_sinkhorn_transport_class():
"""test_sinkhorn_transport
"""
ns = 150
nt = 200
Xs, ys = make_data_classif('3gauss', ns)
Xt, yt = make_data_classif('3gauss2', nt)
otda = ot.da.UnbalancedSinkhornTransport()
# test its computed
otda.fit(Xs=Xs, Xt=Xt)
assert hasattr(otda, "cost_")
assert hasattr(otda, "coupling_")
assert hasattr(otda, "log_")
# test dimensions of coupling
assert_equal(otda.cost_.shape, ((Xs.shape[0], Xt.shape[0])))
assert_equal(otda.coupling_.shape, ((Xs.shape[0], Xt.shape[0])))
# test transform
transp_Xs = otda.transform(Xs=Xs)
assert_equal(transp_Xs.shape, Xs.shape)
Xs_new, _ = make_data_classif('3gauss', ns + 1)
transp_Xs_new = otda.transform(Xs_new)
# check that the oos method is working
assert_equal(transp_Xs_new.shape, Xs_new.shape)
# test inverse transform
transp_Xt = otda.inverse_transform(Xt=Xt)
assert_equal(transp_Xt.shape, Xt.shape)
Xt_new, _ = make_data_classif('3gauss2', nt + 1)
transp_Xt_new = otda.inverse_transform(Xt=Xt_new)
# check that the oos method is working
assert_equal(transp_Xt_new.shape, Xt_new.shape)
# test fit_transform
transp_Xs = otda.fit_transform(Xs=Xs, Xt=Xt)
assert_equal(transp_Xs.shape, Xs.shape)
# test unsupervised vs semi-supervised mode
otda_unsup = ot.da.SinkhornTransport()
otda_unsup.fit(Xs=Xs, Xt=Xt)
n_unsup = np.sum(otda_unsup.cost_)
otda_semi = ot.da.SinkhornTransport()
otda_semi.fit(Xs=Xs, ys=ys, Xt=Xt, yt=yt)
assert_equal(otda_semi.cost_.shape, ((Xs.shape[0], Xt.shape[0])))
n_semisup = np.sum(otda_semi.cost_)
# check that the cost matrix norms are indeed different
assert n_unsup != n_semisup, "semisupervised mode not working"
# check everything runs well with log=True
otda = ot.da.SinkhornTransport(log=True)
otda.fit(Xs=Xs, ys=ys, Xt=Xt)
assert len(otda.log_.keys()) != 0
def _compare(synapses, expected):
conn_matrix = np.zeros((len(synapses.source), len(synapses.target)))
for i, j in zip(synapses.i[:], synapses.j[:]):
conn_matrix[i, j] += 1
assert_equal(conn_matrix, expected)
def test_state_variable_indexing():
G1 = NeuronGroup(5, 'v:1')
G2 = NeuronGroup(7, 'v:1')
S = Synapses(G1, G2, 'w:1')
S.connect(True, n=2)
S.w[:, :, 0] = '5*i + j'
S.w[:, :, 1] = '35 + 5*i + j'
#Slicing
assert len(S.w[:]) == len(S.w[:, :]) == len(S.w[:, :, :]) == len(G1)*len(G2)*2
assert len(S.w[0:]) == len(S.w[0:, 0:]) == len(S.w[0:, 0:, 0:]) == len(G1)*len(G2)*2
assert len(S.w[0]) == len(S.w[0, :]) == len(S.w[0, :, :]) == len(G2)*2
assert len(S.w[0:2]) == len(S.w[0:2, :]) == len(S.w[0:2, :, :]) == 2*len(G2)*2
assert len(S.w[:, 0]) == len(S.w[:, 0, :]) == len(G1)*2
assert len(S.w[:, 0:2]) == len(S.w[:, 0:2, :]) == 2*len(G1)*2
assert len(S.w[:, :2]) == len(S.w[:, :2, :]) == 2*len(G1)*2
assert len(S.w[:, :, 0]) == len(G1)*len(G2)
assert len(S.w[:, :, 0:2]) == len(G1)*len(G2)*2
assert len(S.w[:, :, :2]) == len(G1)*len(G2)*2
#Array-indexing (not yet supported for synapse index)
assert_equal(S.w[0:3], S.w[[0, 1, 2]])
assert_equal(S.w[0:3], S.w[[0, 1, 2], np.arange(len(G2))])
assert_equal(S.w[:, 0:3], S.w[:, [0, 1, 2]])
assert_equal(S.w[:, 0:3], S.w[np.arange(len(G1)), [0, 1, 2]])
#string-based indexing
assert_equal(S.w[0:3], S.w['i<3'])
assert_equal(S.w[:, 0:3], S.w['j<3'])
assert_equal(S.w[:, :, 0], S.w['k==0'])
#invalid indices
assert_raises(IndexError, lambda: S.w.__getitem__((1, 2, 3, 4)))
assert_raises(IndexError, lambda: S.w.__getitem__(object()))
def test_spike_monitor():
G = NeuronGroup(10, 'v:1', threshold='v>1', reset='v=0')
G.v[0] = 1.1
G.v[2] = 1.1
G.v[5] = 1.1
SG = G[3:]
SG2 = G[:3]
s_mon = SpikeMonitor(G)
sub_s_mon = SpikeMonitor(SG)
sub_s_mon2 = SpikeMonitor(SG2)
run(defaultclock.dt)
assert_equal(s_mon.i, np.array([0, 2, 5]))
assert_equal(s_mon.t_, np.zeros(3))
assert_equal(sub_s_mon.i, np.array([2]))
assert_equal(sub_s_mon.t_, np.zeros(1))
assert_equal(sub_s_mon2.i, np.array([0, 2]))
assert_equal(sub_s_mon2.t_, np.zeros(2))
expected = np.zeros(10, dtype=int)
expected[[0, 2, 5]] = 1
assert_equal(s_mon.count, expected)
expected = np.zeros(7, dtype=int)
expected[[2]] = 1
assert_equal(sub_s_mon.count, expected)
assert_equal(sub_s_mon2.count, np.array([1, 0, 1]))
def testHalfSizeParameter():
raw = rawpy.imread(rawTestPath)
s = raw.sizes
rgb = raw.postprocess(half_size=True)
assert_equal(rgb.shape[0], s.height // 2)
assert_equal(rgb.shape[1], s.width // 2)
def test_synapse_access():
G1 = NeuronGroup(10, 'v:1', threshold='False')
G1.v = 'i'
G2 = NeuronGroup(20, 'v:1', threshold='False')
G2.v = 'i'
SG1 = G1[:5]
SG2 = G2[10:]
S = Synapses(SG1, SG2, 'w:1', on_pre='v+=w')
S.connect(True)
S.w['j == 0'] = 5
assert all(S.w['j==0'] == 5)
S.w[2, 2] = 7
assert all(S.w['i==2 and j==2'] == 7)
S.w = '2*j'
assert all(S.w[:, 1] == 2)
assert len(S.w[:, 10]) == 0
assert len(S.w['j==10']) == 0
# Test referencing pre- and postsynaptic variables
assert_equal(S.w[2:, :], S.w['v_pre >= 2'])
assert_equal(S.w[:, :5], S.w['v_post < 15'])
S.w = 'v_post'
assert_equal(S.w[:], S.j[:] + 10)
S.w = 'v_post + v_pre'
assert_equal(S.w[:], S.j[:] + 10 + S.i[:])
# Test using subgroups as indices
assert len(S) == len(S.w[SG1, SG2])
assert_equal(S.w[SG1, 1], S.w[:, 1])
assert_equal(S.w[1, SG2], S.w[1, :])
assert len(S.w[SG1, 10]) == 0
def test_poly2d(self):
p2 = models.Polynomial2D(degree=3)
p2.c0_0 = 5
utils.assert_equal(p2.parameters, [5, 0, 0, 0, 0, 0, 0, 0, 0, 0])
def test_state_variable_access_strings():
for codeobj_class in codeobj_classes:
G = NeuronGroup(10, 'v:volt', codeobj_class=codeobj_class)
G.v = np.arange(10) * volt
# Indexing with strings
assert G.v['i==2'] == G.v[2]
assert G.v_['i==2'] == G.v_[2]
assert_equal(G.v['v >= 3*volt'], G.v[3:])
assert_equal(G.v_['v >= 3*volt'], G.v_[3:])
# Should also check for units
assert_raises(DimensionMismatchError, lambda: G.v['v >= 3'])
assert_raises(DimensionMismatchError, lambda: G.v['v >= 3*second'])
# Setting with strings
# --------------------
# String value referring to i
G.v = '2*i*volt'
assert_equal(G.v[:], 2*np.arange(10)*volt)
# String value referring to i
G.v[:5] = '3*i*volt'
assert_equal(G.v[:],
np.array([0, 3, 6, 9, 12, 10, 12, 14, 16, 18])*volt)
G.v = np.arange(10) * volt
# String value referring to a state variable
G.v = '2*v'
assert_equal(G.v[:], 2*np.arange(10)*volt)
G.v[:5] = '2*v'
assert_equal(G.v[:],
np.array([0, 4, 8, 12, 16, 10, 12, 14, 16, 18])*volt)
G.v = np.arange(10) * volt
# String value referring to state variables, i, and an external variable
ext = 5*volt
G.v = 'v + ext + (N + i)*volt'
assert_equal(G.v[:], 2*np.arange(10)*volt + 15*volt)
G.v = np.arange(10) * volt
G.v[:5] = 'v + ext + (N + i)*volt'
assert_equal(G.v[:],
np.array([15, 17, 19, 21, 23, 5, 6, 7, 8, 9])*volt)
G.v = 'v + randn()*volt' # only check that it doesn't raise an error
G.v[:5] = 'v + randn()*volt' # only check that it doesn't raise an error
G.v = np.arange(10) * volt
# String index using a random number
G.v['rand() <= 1'] = 0*mV
assert_equal(G.v[:], np.zeros(10)*volt)
G.v = np.arange(10) * volt
# String index referring to i and setting to a scalar value
G.v['i>=5'] = 0*mV
assert_equal(G.v[:], np.array([0, 1, 2, 3, 4, 0, 0, 0, 0, 0])*volt)
# String index referring to a state variable
G.v['v<3*volt'] = 0*mV
assert_equal(G.v[:], np.array([0, 0, 0, 3, 4, 0, 0, 0, 0, 0])*volt)
# String index referring to state variables, i, and an external variable
ext = 2*volt
G.v['v>=ext and i==(N-6)'] = 0*mV
assert_equal(G.v[:], np.array([0, 0, 0, 3, 0, 0, 0, 0, 0, 0])*volt)
G.v = np.arange(10) * volt
# Strings for both condition and values
G.v['i>=5'] = 'v*2'
assert_equal(G.v[:], np.array([0, 1, 2, 3, 4, 10, 12, 14, 16, 18])*volt)
G.v['v>=5*volt'] = 'i*volt'
assert_equal(G.v[:], np.arange(10)*volt)
def test_subexpression_with_constant():
g = 2
G = NeuronGroup(1, '''I = 1*g : 1''')
assert_equal(G.I[:], np.array([2]))
def test_state_variable_access():
for codeobj_class in codeobj_classes:
G = NeuronGroup(10, 'v:volt', codeobj_class=codeobj_class)
G.v = np.arange(10) * volt
assert_equal(np.asarray(G.v[:]), np.arange(10))
assert have_same_dimensions(G.v[:], volt)
assert_equal(np.asarray(G.v[:]), G.v_[:])
# Accessing single elements, slices and arrays
assert G.v[5] == 5 * volt
assert G.v_[5] == 5
assert_equal(G.v[:5], np.arange(5) * volt)
assert_equal(G.v_[:5], np.arange(5))
assert_equal(G.v[[0, 5]], [0, 5] * volt)
assert_equal(G.v_[[0, 5]], np.array([0, 5]))
# Illegal indexing
assert_raises(IndexError, lambda: G.v[0, 0])
assert_raises(IndexError, lambda: G.v_[0, 0])
assert_raises(TypeError, lambda: G.v[object()])
assert_raises(TypeError, lambda: G.v_[object()])
# A string representation should not raise any error
assert len(str(G.v))
assert len(repr(G.v))
assert len(str(G.v_))
assert len(repr(G.v_))
def test_state_variables():
'''
Test the setting and accessing of state variables.
'''
for codeobj_class in codeobj_classes:
G = NeuronGroup(10, 'v : volt', codeobj_class=codeobj_class)
# The variable N should be always present
assert G.N == 10
# But it should be read-only
assert_raises(TypeError, lambda: G.__setattr__('N', 20))
assert_raises(TypeError, lambda: G.__setattr__('N_', 20))
G.v = -70*mV
assert_raises(DimensionMismatchError, lambda: G.__setattr__('v', -70))
G.v_ = float(-70*mV)
assert_allclose(G.v[:], -70*mV)
G.v = -70*mV + np.arange(10)*mV
assert_allclose(G.v[:], -70*mV + np.arange(10)*mV)
G.v = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9] * volt
assert_allclose(G.v[:], np.arange(10) * volt)
# incorrect size
assert_raises(ValueError, lambda: G.__setattr__('v', [0, 1]*volt))
assert_raises(ValueError, lambda: G.__setattr__('v', np.arange(11)*volt))
G.v = -70*mV
# Numpy methods should be able to deal with state variables
# (discarding units)
assert_allclose(np.mean(G.v), float(-70*mV))
# Getting the content should return a Quantity object which then natively
# supports numpy functions that access a method
assert_allclose(np.mean(G.v[:]), -70*mV)
# You should also be able to set variables with a string
G.v = '-70*mV + i*mV'
assert_allclose(G.v[0], -70*mV)
assert_allclose(G.v[9], -61*mV)
assert_allclose(G.v[:], -70*mV + np.arange(10)*mV)
# And it should raise an unit error if the units are incorrect
assert_raises(DimensionMismatchError,
lambda: G.__setattr__('v', '70 + i'))
assert_raises(DimensionMismatchError,
lambda: G.__setattr__('v', '70 + i*mV'))
# Calculating with state variables should work too
# With units
assert all(G.v - G.v == 0)
assert all(G.v - G.v[:] == 0*mV)
assert all(G.v[:] - G.v == 0*mV)
assert all(G.v + 70*mV == G.v[:] + 70*mV)
assert all(70*mV + G.v == G.v[:] + 70*mV)
assert all(G.v + G.v == 2*G.v)
assert all(G.v / 2.0 == 0.5*G.v)
assert all(1.0 / G.v == 1.0 / G.v[:])
assert_equal((-G.v)[:], -G.v[:])
assert_equal((+G.v)[:], G.v[:])
#Without units
assert all(G.v_ - G.v_ == 0)
assert all(G.v_ - G.v_[:] == 0)
assert all(G.v_[:] - G.v_ == 0)
assert all(G.v_ + float(70*mV) == G.v_[:] + float(70*mV))
assert all(float(70*mV) + G.v_ == G.v_[:] + float(70*mV))
assert all(G.v_ + G.v_ == 2*G.v_)
assert all(G.v_ / 2.0 == 0.5*G.v_)
assert all(1.0 / G.v_ == 1.0 / G.v_[:])
assert_equal((-G.v)[:], -G.v[:])
assert_equal((+G.v)[:], G.v[:])
# And in-place modification should work as well
G.v += 10*mV
G.v -= 10*mV
G.v *= 2
G.v /= 2.0
# with unit checking
assert_raises(DimensionMismatchError, lambda: G.v.__iadd__(3*second))
assert_raises(DimensionMismatchError, lambda: G.v.__iadd__(3))
assert_raises(DimensionMismatchError, lambda: G.v.__imul__(3*second))
# in-place modification with strings should not work
assert_raises(TypeError, lambda: G.v.__iadd__('string'))
assert_raises(TypeError, lambda: G.v.__imul__('string'))
assert_raises(TypeError, lambda: G.v.__idiv__('string'))
assert_raises(TypeError, lambda: G.v.__isub__('string'))
def compare_faster():
variables = collections.OrderedDict()
variables['alpha'] = dict(min=-180,
max=180,
description="angle",
resolution=5,
units='deg',
units_display='deg')
variables['r'] = dict(min=3,
max=5,
description="distance",
resolution=0.1,
units='m',
units_display='cm')
# this will fail if precision is float32
gh = GridHelper(variables, precision='float64')
val_fast = gh.create_new()
val_fast.fill(0)
val_slow = gh.create_new()
val_slow.fill(0)
od = dtu.get_output_dir_for_test()
F = 1
alpha0 = 7
# r0 = 4
r0 = 4.1
w0 = 1.
value = dict(alpha=alpha0, r=r0)
gh.add_vote(val_slow, value, w0, F)
assert_equal(np.sum(val_slow > 0), 9)
values = np.zeros((2, 1))
values[0, 0] = alpha0
values[1, 0] = r0
weights = np.zeros(1)
weights[0] = w0
gh.add_vote_faster(val_fast, values, weights, F)
assert_equal(np.sum(val_fast > 0), 9)
d = grid_helper_plot(gh, val_slow)
fn = os.path.join(od, 'compare_faster_slow.jpg')
dtu.write_data_to_file(d.get_png(), fn)
d = grid_helper_plot(gh, val_fast)
fn = os.path.join(od, 'compare_faster_fast.jpg')
dtu.write_data_to_file(d.get_png(), fn)
D = val_fast - val_slow
diff = np.max(np.abs(D))
print('diff: %r' % diff)
if diff > 1e-8:
print(dtu.indent(array_as_string_sign(val_fast), 'val_fast '))
print(dtu.indent(array_as_string_sign(val_slow), 'val_slow '))
print(dtu.indent(array_as_string_sign(D), 'Diff '))
print('non zero val_fast: %s' % val_fast[val_fast > 0])
print('non zero val_slow: %s' % val_slow[val_slow > 0])
assert_almost_equal(val_fast, val_slow)
def test_synapse_creation_generator_multiple_synapses():
G1 = NeuronGroup(10, 'v:1', threshold='False')
G2 = NeuronGroup(20, 'v:1', threshold='False')
G1.v = 'i'
G2.v = '10 + i'
SG1 = G1[:5]
SG2 = G2[10:]
S1 = Synapses(SG1, SG2, 'w:1', on_pre='v+=w')
S1.connect(j='k for k in range(N_post)', n='i')
S2 = Synapses(SG1, SG2, 'w:1', on_pre='v+=w')
S2.connect(j='k for k in range(N_post)', n='j')
S3 = Synapses(SG2, SG1, 'w:1', on_pre='v+=w')
S3.connect(j='k for k in range(N_post)', n='i')
S4 = Synapses(SG2, SG1, 'w:1', on_pre='v+=w')
S4.connect(j='k for k in range(N_post)', n='j')
S5 = Synapses(SG1, SG2, 'w:1', on_pre='v+=w')
S5.connect(j='k for k in range(N_post)', n='i+j')
S6 = Synapses(SG2, SG1, 'w:1', on_pre='v+=w')
S6.connect(j='k for k in range(N_post)', n='i+j')
S7 = Synapses(SG1, SG2, 'w:1', on_pre='v+=w')
S7.connect(j='k for k in range(N_post)', n='int(v_pre>2)*2')
S8 = Synapses(SG2, SG1, 'w:1', on_pre='v+=w')
S8.connect(j='k for k in range(N_post)', n='int(v_post>2)*2')
S9 = Synapses(SG1, SG2, 'w:1', on_pre='v+=w')
S9.connect(j='k for k in range(N_post)', n='int(v_post>22)*2')
S10 = Synapses(SG2, SG1, 'w:1', on_pre='v+=w')
S10.connect(j='k for k in range(N_post)', n='int(v_pre>22)*2')
run(0*ms) # for standalone
# straightforward loop instead of doing something clever...
for source in xrange(len(SG1)):
assert_equal(S1.j[source, :], np.arange(len(SG2)).repeat(source))
assert_equal(S2.j[source, :], np.arange(len(SG2)).repeat(np.arange(len(SG2))))
assert_equal(S3.i[:, source], np.arange(len(SG2)).repeat(np.arange(len(SG2))))
assert_equal(S4.i[:, source], np.arange(len(SG2)).repeat(source))
assert_equal(S5.j[source, :], np.arange(len(SG2)).repeat(np.arange(len(SG2))+source))
assert_equal(S6.i[:, source], np.arange(len(SG2)).repeat(np.arange(len(SG2)) + source))
if source > 2:
assert_equal(S7.j[source, :], np.arange(len(SG2)).repeat(2))
assert_equal(S8.i[:, source], np.arange(len(SG2)).repeat(2))
else:
assert len(S7.j[source, :]) == 0
assert len(S8.i[:, source]) == 0
assert_equal(S9.j[source, :], np.arange(3, len(SG2)).repeat(2))
assert_equal(S10.i[:, source], np.arange(3, len(SG2)).repeat(2))
def test_subexpression_references():
'''
Assure that subexpressions in targeted groups are handled correctly.
'''
G = NeuronGroup(10, '''v : 1
v2 = 2*v : 1''')
G.v = np.arange(10)
SG1 = G[:5]
SG2 = G[5:]
S1 = Synapses(SG1, SG2, '''w : 1
u = v2_post + 1 : 1
x = v2_pre + 1 : 1''')
S1.connect('i==(5-1-j)')
assert_equal(S1.i[:], np.arange(5))
assert_equal(S1.j[:], np.arange(5)[::-1])
assert_equal(S1.u[:], np.arange(10)[:-6:-1]*2+1)
assert_equal(S1.x[:], np.arange(5)*2+1)
S2 = Synapses(G, SG2, '''w : 1
u = v2_post + 1 : 1
x = v2_pre + 1 : 1''')
S2.connect('i==(5-1-j)')
assert_equal(S2.i[:], np.arange(5))
assert_equal(S2.j[:], np.arange(5)[::-1])
assert_equal(S2.u[:], np.arange(10)[:-6:-1]*2+1)
assert_equal(S2.x[:], np.arange(5)*2+1)
S3 = Synapses(SG1, G, '''w : 1
u = v2_post + 1 : 1
x = v2_pre + 1 : 1''')
S3.connect('i==(10-1-j)')
assert_equal(S3.i[:], np.arange(5))
assert_equal(S3.j[:], np.arange(10)[:-6:-1])
assert_equal(S3.u[:], np.arange(10)[:-6:-1]*2+1)
assert_equal(S3.x[:], np.arange(5)*2+1)
def test_conditional_gaussian_dependency_matrix(self):
length = 100
n_samples = 1000
X = array([sample_markov_chain(length) for _ in range(n_samples)])
# Next two should be equal
s0 = AnomalyDetector(
P_ConditionalGaussianDependencyMatrix(
range(length), length)).fit(X).anomaly_score(X)
ad1 = AnomalyDetector(
P_ConditionalGaussianCombiner([
P_ConditionalGaussian([i + 1], [i]) for i in range(length - 1)
] + [P_ConditionalGaussian([0], [])]), cr_plus).fit(X)
s1 = ad1.anomaly_score(X)
assert_allclose(s0, s1, rtol=0.0001) # OK
# Most likely, these two are not equal but highly correlated
ad2 = AnomalyDetector(
[P_ConditionalGaussian([i], []) for i in range(length)],
cr_plus).fit(X)
s2 = ad2.anomaly_score(X)
ad3 = AnomalyDetector(
P_ConditionalGaussianCombiner(
[P_ConditionalGaussian([i], []) for i in range(length)]),
cr_plus).fit(X)
s3 = ad3.anomaly_score(X)
assert_equal(pearsonr(s2, s3) > 0.985, True)
# Test classification
Y = array([sample_markov_chain(length, 0.2) for _ in range(n_samples)])
Z = array([sample_markov_chain(length, 0.3) for _ in range(n_samples)])
data = r_[X, Y, Z]
labels = r_[['X'] * len(X), ['Y'] * len(Y), ['Z'] * len(Z)]
data_index = shuffle(range(len(data)))
training_set = data_index[:n_samples * 2]
test_set = data_index[n_samples * 2:]
models = {
'independent gaussian':
AnomalyDetector([P_Gaussian([i]) for i in range(length)], cr_plus),
'independent conditional gaussian':
AnomalyDetector(
[P_ConditionalGaussian([i], []) for i in range(length)],
cr_plus),
'independent conditional gaussian with combiner':
AnomalyDetector(
P_ConditionalGaussianCombiner(
[P_ConditionalGaussian([i], []) for i in range(length)])),
'single conditional gaussian with combiner':
AnomalyDetector(
P_ConditionalGaussianCombiner([
P_ConditionalGaussian([i], [i - 1])
for i in range(1, length)
] + [P_ConditionalGaussian([0], [])])),
'dependency matrix':
AnomalyDetector(
P_ConditionalGaussianDependencyMatrix(range(length), length))
}
all_acc = {}
for key in models:
ad = models[key].fit(data[training_set], labels[training_set])
adclf = SklearnClassifier.clf(ad)
labels_predicted = adclf.predict(data[test_set])
accuracy = sum(labels[test_set] == labels_predicted) / float(
len(test_set))
all_acc[key] = accuracy
print key, "accuracy = ", accuracy
assert_close(all_acc['independent gaussian'],
all_acc['independent conditional gaussian'],
decimal=2)
assert_close(all_acc['independent gaussian'],
all_acc['independent conditional gaussian with combiner'],
decimal=2)
assert_close(all_acc['single conditional gaussian with combiner'],
all_acc['dependency matrix'],
decimal=2)
def test_rate_arrays():
P = PoissonGroup(2, np.array([0, 1. / defaultclock.dt]) * Hz)
spikes = SpikeMonitor(P)
run(2 * defaultclock.dt)
assert_equal(spikes.count, np.array([0, 2]))
def test_synapses_access_subgroups():
G1 = NeuronGroup(5, 'x:1')
G2 = NeuronGroup(10, 'y:1')
SG1 = G1[2:5]
SG2 = G2[4:9]
S = Synapses(G1, G2, 'w:1')
S.connect()
S.w[SG1, SG2] = 1
assert_equal(S.w['(i>=2 and i<5) and (j>=4 and j<9)'], 1)
assert_equal(S.w['not ((i>=2 and i<5) and (j>=4 and j<9))'], 0)
S.w = 0
S.w[SG1, :] = 1
assert_equal(S.w['i>=2 and i<5'], 1)
assert_equal(S.w['not (i>=2 and i<5)'], 0)
S.w = 0
S.w[:, SG2] = 1
assert_equal(S.w['j>=4 and j<9'], 1)
assert_equal(S.w['not (j>=4 and j<9)'], 0)
def test_sinkhorn_l1l2_transport_class():
"""test_sinkhorn_transport
"""
ns = 150
nt = 200
Xs, ys = make_data_classif('3gauss', ns)
Xt, yt = make_data_classif('3gauss2', nt)
otda = ot.da.SinkhornL1l2Transport()
# test its computed
otda.fit(Xs=Xs, ys=ys, Xt=Xt)
assert hasattr(otda, "cost_")
assert hasattr(otda, "coupling_")
assert hasattr(otda, "log_")
# test dimensions of coupling
assert_equal(otda.cost_.shape, ((Xs.shape[0], Xt.shape[0])))
assert_equal(otda.coupling_.shape, ((Xs.shape[0], Xt.shape[0])))
# test margin constraints
mu_s = unif(ns)
mu_t = unif(nt)
assert_allclose(
np.sum(otda.coupling_, axis=0), mu_t, rtol=1e-3, atol=1e-3)
assert_allclose(
np.sum(otda.coupling_, axis=1), mu_s, rtol=1e-3, atol=1e-3)
# test transform
transp_Xs = otda.transform(Xs=Xs)
assert_equal(transp_Xs.shape, Xs.shape)
Xs_new, _ = make_data_classif('3gauss', ns + 1)
transp_Xs_new = otda.transform(Xs_new)
# check that the oos method is working
assert_equal(transp_Xs_new.shape, Xs_new.shape)
# test inverse transform
transp_Xt = otda.inverse_transform(Xt=Xt)
assert_equal(transp_Xt.shape, Xt.shape)
Xt_new, _ = make_data_classif('3gauss2', nt + 1)
transp_Xt_new = otda.inverse_transform(Xt=Xt_new)
# check that the oos method is working
assert_equal(transp_Xt_new.shape, Xt_new.shape)
# test fit_transform
transp_Xs = otda.fit_transform(Xs=Xs, ys=ys, Xt=Xt)
assert_equal(transp_Xs.shape, Xs.shape)
# test unsupervised vs semi-supervised mode
otda_unsup = ot.da.SinkhornL1l2Transport()
otda_unsup.fit(Xs=Xs, ys=ys, Xt=Xt)
n_unsup = np.sum(otda_unsup.cost_)
otda_semi = ot.da.SinkhornL1l2Transport()
otda_semi.fit(Xs=Xs, ys=ys, Xt=Xt, yt=yt)
assert_equal(otda_semi.cost_.shape, ((Xs.shape[0], Xt.shape[0])))
n_semisup = np.sum(otda_semi.cost_)
# check that the cost matrix norms are indeed different
assert n_unsup != n_semisup, "semisupervised mode not working"
# check that the coupling forbids mass transport between labeled source
# and labeled target samples
mass_semi = np.sum(
otda_semi.coupling_[otda_semi.cost_ == otda_semi.limit_max])
mass_semi = otda_semi.coupling_[otda_semi.cost_ == otda_semi.limit_max]
assert_allclose(mass_semi, np.zeros_like(mass_semi),
rtol=1e-9, atol=1e-9)
# check everything runs well with log=True
otda = ot.da.SinkhornL1l2Transport(log=True)
otda.fit(Xs=Xs, ys=ys, Xt=Xt)
assert len(otda.log_.keys()) != 0
def test_subexpression_no_references():
'''
Assure that subexpressions are handled correctly, even
when the subgroups are created on-the-fly.
'''
G = NeuronGroup(10, '''v : 1
v2 = 2*v : 1''')
G.v = np.arange(10)
assert_equal(G[5:].v2, np.arange(5, 10)*2)
S1 = Synapses(G[:5], G[5:], '''w : 1
u = v2_post + 1 : 1
x = v2_pre + 1 : 1''')
S1.connect('i==(5-1-j)')
assert_equal(S1.i[:], np.arange(5))
assert_equal(S1.j[:], np.arange(5)[::-1])
assert_equal(S1.u[:], np.arange(10)[:-6:-1]*2+1)
assert_equal(S1.x[:], np.arange(5)*2+1)
S2 = Synapses(G, G[5:], '''w : 1
u = v2_post + 1 : 1
x = v2_pre + 1 : 1''')
S2.connect('i==(5-1-j)')
assert_equal(S2.i[:], np.arange(5))
assert_equal(S2.j[:], np.arange(5)[::-1])
assert_equal(S2.u[:], np.arange(10)[:-6:-1]*2+1)
assert_equal(S2.x[:], np.arange(5)*2+1)
S3 = Synapses(G[:5], G, '''w : 1
u = v2_post + 1 : 1
x = v2_pre + 1 : 1''')
S3.connect('i==(10-1-j)')
assert_equal(S3.i[:], np.arange(5))
assert_equal(S3.j[:], np.arange(10)[:-6:-1])
assert_equal(S3.u[:], np.arange(10)[:-6:-1]*2+1)
assert_equal(S3.x[:], np.arange(5)*2+1)
def assert_equal(actual, desired, err_msg='', verbose=True):
"""
Alternative naming for assertArrayEqual.
"""
return nptu.assert_equal(actual, desired, err_msg=err_msg, verbose=verbose)
def test_alternative_indexing():
G = NeuronGroup(10, 'v : integer')
G.v = 'i'
assert_equal(G[-3:].v, np.array([7, 8, 9]))
assert_equal(G[3].v, np.array([3]))
assert_equal(G[[3, 4, 5]].v, np.array([3, 4, 5]))
def test_mapping_transport_class():
"""test_mapping_transport
"""
ns = 60
nt = 120
Xs, ys = make_data_classif('3gauss', ns)
Xt, yt = make_data_classif('3gauss2', nt)
Xs_new, _ = make_data_classif('3gauss', ns + 1)
##########################################################################
# kernel == linear mapping tests
##########################################################################
# check computation and dimensions if bias == False
otda = ot.da.MappingTransport(kernel="linear", bias=False)
otda.fit(Xs=Xs, Xt=Xt)
assert hasattr(otda, "coupling_")
assert hasattr(otda, "mapping_")
assert hasattr(otda, "log_")
assert_equal(otda.coupling_.shape, ((Xs.shape[0], Xt.shape[0])))
assert_equal(otda.mapping_.shape, ((Xs.shape[1], Xt.shape[1])))
# test margin constraints
mu_s = unif(ns)
mu_t = unif(nt)
assert_allclose(
np.sum(otda.coupling_, axis=0), mu_t, rtol=1e-3, atol=1e-3)
assert_allclose(
np.sum(otda.coupling_, axis=1), mu_s, rtol=1e-3, atol=1e-3)
# test transform
transp_Xs = otda.transform(Xs=Xs)
assert_equal(transp_Xs.shape, Xs.shape)
transp_Xs_new = otda.transform(Xs_new)
# check that the oos method is working
assert_equal(transp_Xs_new.shape, Xs_new.shape)
# check computation and dimensions if bias == True
otda = ot.da.MappingTransport(kernel="linear", bias=True)
otda.fit(Xs=Xs, Xt=Xt)
assert_equal(otda.coupling_.shape, ((Xs.shape[0], Xt.shape[0])))
assert_equal(otda.mapping_.shape, ((Xs.shape[1] + 1, Xt.shape[1])))
# test margin constraints
mu_s = unif(ns)
mu_t = unif(nt)
assert_allclose(
np.sum(otda.coupling_, axis=0), mu_t, rtol=1e-3, atol=1e-3)
assert_allclose(
np.sum(otda.coupling_, axis=1), mu_s, rtol=1e-3, atol=1e-3)
# test transform
transp_Xs = otda.transform(Xs=Xs)
assert_equal(transp_Xs.shape, Xs.shape)
transp_Xs_new = otda.transform(Xs_new)
# check that the oos method is working
assert_equal(transp_Xs_new.shape, Xs_new.shape)
##########################################################################
# kernel == gaussian mapping tests
##########################################################################
# check computation and dimensions if bias == False
otda = ot.da.MappingTransport(kernel="gaussian", bias=False)
otda.fit(Xs=Xs, Xt=Xt)
assert_equal(otda.coupling_.shape, ((Xs.shape[0], Xt.shape[0])))
assert_equal(otda.mapping_.shape, ((Xs.shape[0], Xt.shape[1])))
# test margin constraints
mu_s = unif(ns)
mu_t = unif(nt)
assert_allclose(
np.sum(otda.coupling_, axis=0), mu_t, rtol=1e-3, atol=1e-3)
assert_allclose(
np.sum(otda.coupling_, axis=1), mu_s, rtol=1e-3, atol=1e-3)
# test transform
transp_Xs = otda.transform(Xs=Xs)
assert_equal(transp_Xs.shape, Xs.shape)
transp_Xs_new = otda.transform(Xs_new)
# check that the oos method is working
assert_equal(transp_Xs_new.shape, Xs_new.shape)
# check computation and dimensions if bias == True
otda = ot.da.MappingTransport(kernel="gaussian", bias=True)
otda.fit(Xs=Xs, Xt=Xt)
assert_equal(otda.coupling_.shape, ((Xs.shape[0], Xt.shape[0])))
assert_equal(otda.mapping_.shape, ((Xs.shape[0] + 1, Xt.shape[1])))
# test margin constraints
mu_s = unif(ns)
mu_t = unif(nt)
assert_allclose(
np.sum(otda.coupling_, axis=0), mu_t, rtol=1e-3, atol=1e-3)
assert_allclose(
np.sum(otda.coupling_, axis=1), mu_s, rtol=1e-3, atol=1e-3)
# test transform
transp_Xs = otda.transform(Xs=Xs)
assert_equal(transp_Xs.shape, Xs.shape)
transp_Xs_new = otda.transform(Xs_new)
# check that the oos method is working
assert_equal(transp_Xs_new.shape, Xs_new.shape)
# check everything runs well with log=True
otda = ot.da.MappingTransport(kernel="gaussian", log=True)
otda.fit(Xs=Xs, Xt=Xt)
assert len(otda.log_.keys()) != 0
def testPolynomial1D(self):
d = {'c0': 11, 'c1': 12, 'c2': 13, 'c3': 14}
p1 = models.Polynomial1D(3, **d)
utils.assert_equal(p1.parameters, [11, 12, 13, 14])
def testarith(self):
adder = 3
array_adder = array([1, 2, 3])
# add
b = self.mix + adder
self.assertIsInstance(b, OrderedDPMixture,
'integer addition return wrong type')
assert_equal(b.mus[0], self.mix.mus[0] + adder,
'integer addition returned wrong value')
c = self.mix + array_adder
self.assertIsInstance(c, OrderedDPMixture,
'array addition return wrong type')
assert_array_equal(c.mus[0], self.mix.mus[0] + array_adder,
'array addition returned wrong value')
# radd
b = adder + self.mix
self.assertIsInstance(b, OrderedDPMixture,
'integer addition return wrong type')
assert_array_equal(b.mus[0], adder + self.mix.mus[0],
'integer addition returned wrong value')
c = array_adder + self.mix
self.assertIsInstance(c, OrderedDPMixture,
'array addition return wrong type')
assert_array_equal(c.mus[0], array_adder + self.mix.mus[0],
'array addition returned wrong value')
# sub
b = self.mix - adder
self.assertIsInstance(b, OrderedDPMixture,
'integer subtraction return wrong type')
assert_array_equal(b.mus[0], self.mix.mus[0] - adder,
'integer subtraction returned wrong value')
c = self.mix - array_adder
self.assertIsInstance(c, OrderedDPMixture,
'array subtraction return wrong type')
assert_array_equal(c.mus[0], self.mix.mus[0] - array_adder,
'array subtraction returned wrong value')
# rsub
b = adder - self.mix
self.assertIsInstance(b, OrderedDPMixture,
'integer subtraction return wrong type')
assert_array_equal(b.mus[0], adder - self.mix.mus[0],
'integer subtraction returned wrong value')
c = array_adder - self.mix
self.assertIsInstance(c, OrderedDPMixture,
'array subtraction return wrong type')
assert_array_equal(c.mus[0], array_adder - self.mix.mus[0],
'array subtraction returned wrong value')
# mul
b = self.mix * adder
self.assertIsInstance(b, OrderedDPMixture,
'integer multiplication return wrong type')
assert_array_equal(b.mus[0], self.mix.mus[0] * adder,
'integer multiplication returned wrong value')
c = self.mix * array_adder
self.assertIsInstance(c, OrderedDPMixture,
'array multiplicaton return wrong type')
assert_array_equal(c.mus[0], dot(self.mix.mus[0], array_adder),
'array multiplication returned wrong value')
# rmul
b = adder * self.mix
self.assertIsInstance(b, OrderedDPMixture,
'integer multiplication return wrong type')
assert_array_equal(b.mus[0], adder * self.mix.mus[0],
'integer multiplication returned wrong value')
c = array_adder * self.mix
self.assertIsInstance(c, OrderedDPMixture,
'array multiplication return wrong type')
assert_array_equal(c.mus[0], dot(array_adder, self.mix.mus[0]),
'array multiplication returned wrong value')
def test_timedarray_no_units():
ta = TimedArray(np.arange(10), dt=0.1 * ms)
G = NeuronGroup(1, 'value = ta(t) + 1: 1', dt=0.1 * ms)
mon = StateMonitor(G, 'value', record=True, dt=0.1 * ms)
run(1.1 * ms)
assert_equal(mon[0].value_, np.clip(np.arange(len(mon[0].t)), 0, 9) + 1)
def testgetitem(self):
assert_equal(self.mu1, self.mix[0].mu, 'getitem failed')
self.mix[0] = self.clust2
assert_equal(self.mu2, self.mix[0].mu, 'getitem failed')
self.mix[0] = self.clust1