def test_KCSD_1D_zero_csd(self):
"""if measured potential is 0, the calculated CSD should be 0"""
params = {'n_sources': 20, 'dist_density': 20,
'source_type': 'gauss_lim'}
k_zero = KCSD(elec_pos=np.array([[1.0], [2.0], [3.0], [4.0], [5.0]]),
sampled_pots=np.array([[0.0], [0.0], [0.0], [0.0], [0.0]]),
params=params)
k_zero.estimate_csd()
self.assertAlmostEqual(norm(k_zero.solver.estimated_csd), 0.0, places=10)
def csd_est_using_kCSD(probe_loc, lfp_bar, lambd=0.0, sigma = sigma):
elec_pos = np.expand_dims(probe_loc, axis=1)
pots = np.expand_dims(lfp_bar, axis=1)
params = {'gdX': np.diff(probe_loc)[0], 'gdY': np.diff(probe_loc)[0], sigma:sigma, lambd:lambd }
k = KCSD(elec_pos, pots, params)
k.solver.lambd = lambd
k.estimate_pots()
k.estimate_csd()
return k.solver.space_X.ravel(), k.solver.estimated_csd.ravel()*sigma
def test_KCSD_1D_csd_reconstruction(self):
"""reconstructed csd should be similar to model csd"""
params = {'sigma': self.sigma, 'source_type': 'gauss_lim',
'x_min': self.xmin, 'x_max': self.xmax, 'h': self.h}
k = KCSD(self.elec_pos, self.meas_pot, params)
k.estimate_csd()
"""print np.max(k.estimated_csd)
plot(k.estimated_csd)
plot(self.true_csd)
show()
print k.X_src"""
for estimated_csd, true_csd in zip(k.solver.estimated_csd, self.true_csd):
self.assertAlmostEqual(estimated_csd, true_csd, places=1)
def test_KCSD_1D_zero_pot(self):
"""if measured potential is 0, the calculated potential should be 0"""
params = {
'n_sources': 20,
'dist_density': 20,
'source_type': 'gauss_lim'
}
k_zero = KCSD(elec_pos=np.array([[1.0], [2.0], [3.0], [4.0], [5.0]]),
sampled_pots=np.array([[0.0, 0.0, 0.0, 0.0, 0.0]]).T,
params=params)
k_zero.estimate_pots()
self.assertAlmostEqual(norm(k_zero.solver.estimated_pots),
0.0,
places=10)
def test_KCSD_1D_pot_reconstruction(self):
"""reconstructed pots should be similar to model pots"""
params = {'sigma': self.sigma, 'source_type': 'gauss_lim',
'x_min': self.xmin, 'x_max': self.xmax,
'h': self.h, 'n_src': 20}
k = KCSD(self.elec_pos, self.meas_pot, params)
k.estimate_pots()
"""print np.max(k.estimated_pots)
plot(k.space_X, k.estimated_pots)
plot(self.x, self.true_pots)
scatter(k.elec_pos, k.sampled_pots)
show()"""
for estimated_pot, true_pot in zip(k.solver.estimated_pots, self.true_pots):
self.assertAlmostEqual(estimated_pot, true_pot, places=1)
def test_KCSD2D_zero_csd(self):
elec_pos = np.array([[0, 0], [0, 1], [1, 0], [1, 1]])
pots = np.array([[0], [0], [0], [0]])
k = KCSD(elec_pos, pots)
k.solver.estimate_csd()
for csd in k.solver.estimated_csd.flatten():
self.assertAlmostEqual(csd, 0.0, places=5)
def test_KCSD_2D_zero_pot(self):
elec_pos = np.array([[0, 0], [0, 1], [1, 0], [1, 1]])
pots = np.array([[0], [0], [0], [0]])
k = KCSD(elec_pos, pots)
k.solver.estimate_pots()
for pot in k.solver.estimated_pots.flatten():
self.assertAlmostEqual(pot, 0.0, places=5)
def test_KCSD_1D_cross_validation_two_electrodes(self):
"""cross validation should promote high lambdas in this case"""
params = {'x_min': 0.0, 'x_max': 1.0, 'gdX': 0.1,
'source_type': 'gauss_lim'}
k = KCSD(elec_pos=np.array([[0.2], [0.7]]),
sampled_pots=np.array([[1.0, 0.5]]).T,
params=params)
k.estimate_pots()
lambdas = np.array([0.1, 0.5, 1.0])
n_elec = k.solver.elec_pos.shape[0]
index_generator = LeaveOneOut(n_elec, indices=True)
k.solver.lambd = cv.choose_lambda(lambdas, k.solver.sampled_pots,
k.solver.k_pot, k.solver.elec_pos,
index_generator)
self.assertEquals(k.solver.lambd, 1.0)
def test_KCSD_3D_zero_pot(self):
"""if the input pots are zero, estimated pots and csd should be zero"""
elec_pos = np.array([(0, 0, 0), (0, 0, 1), (0, 1, 0), (1, 0, 0),
(0, 1, 1), (1, 1, 0), (1, 0, 1), (1, 1, 1)])
pots = np.array([[0], [0], [0], [0], [0], [0], [0], [0]])
params = {
'gdX': 0.2,
'gdY': 0.2,
'gdZ': 0.2,
'n_src': 10
}
k = KCSD(elec_pos, pots, params)
k.estimate_pots()
k.estimate_csd()
for pot in k.solver.estimated_pots.flatten():
self.assertAlmostEqual(pot, 0.0, places=5)
for csd in k.solver.estimated_csd.flatten():
self.assertAlmostEqual(csd, 0.0, places=5)
def test_KCSD_1D_cross_validation_two_electrodes(self):
"""cross validation should promote high lambdas in this case"""
params = {
'x_min': 0.0,
'x_max': 1.0,
'gdX': 0.1,
'source_type': 'gauss_lim'
}
k = KCSD(elec_pos=np.array([[0.2], [0.7]]),
sampled_pots=np.array([[1.0, 0.5]]).T,
params=params)
k.estimate_pots()
lambdas = np.array([0.1, 0.5, 1.0])
n_elec = k.solver.elec_pos.shape[0]
index_generator = LeaveOneOut(n_elec) #, indices=True)
k.solver.lambd = cv.choose_lambda(lambdas, k.solver.sampled_pots,
k.solver.k_pot, k.solver.elec_pos,
index_generator)
self.assertEqual(k.solver.lambd, 1.0)
def test_KCSD_1D_csd_reconstruction(self):
"""reconstructed csd should be similar to model csd"""
params = {
'sigma': self.sigma,
'source_type': 'gauss_lim',
'x_min': self.xmin,
'x_max': self.xmax,
'h': self.h
}
k = KCSD(self.elec_pos, self.meas_pot, params)
k.estimate_csd()
"""print np.max(k.estimated_csd)
plot(k.estimated_csd)
plot(self.true_csd)
show()
print k.X_src"""
for estimated_csd, true_csd in zip(k.solver.estimated_csd,
self.true_csd):
self.assertAlmostEqual(estimated_csd, true_csd, places=1)
def test_KCSD_1D_pot_reconstruction(self):
"""reconstructed pots should be similar to model pots"""
params = {
'sigma': self.sigma,
'source_type': 'gauss_lim',
'x_min': self.xmin,
'x_max': self.xmax,
'h': self.h,
'n_src': 20
}
k = KCSD(self.elec_pos, self.meas_pot, params)
k.estimate_pots()
"""print np.max(k.estimated_pots)
plot(k.space_X, k.estimated_pots)
plot(self.x, self.true_pots)
scatter(k.elec_pos, k.sampled_pots)
show()"""
for estimated_pot, true_pot in zip(k.solver.estimated_pots,
self.true_pots):
self.assertAlmostEqual(estimated_pot, true_pot, places=1)
def test_KCSD_1D_lambda_choice(self):
"""for potentials calculated from model, lambda < 1.0"""
params = {
'sigma': self.sigma,
'source_type': 'gauss_lim',
'x_min': -5.0,
'x_max': 10.0,
'h': self.h
}
k = KCSD(self.elec_pos, self.meas_pot, params)
lambdas = np.array([100.0 / 2**n for n in range(1, 20)])
n_elec = k.solver.elec_pos.shape[0]
index_generator = LeaveOneOut(n_elec) #, indices=True)
k.solver.lambd = cv.choose_lambda(lambdas, k.solver.sampled_pots,
k.solver.k_pot, k.solver.elec_pos,
index_generator)
self.assertLess(k.solver.lambd, 1.0)
def test_KCSD_3D_zero_pot(self):
"""if the input pots are zero, estimated pots and csd should be zero"""
elec_pos = np.array([(0, 0, 0), (0, 0, 1), (0, 1, 0), (1, 0, 0),
(0, 1, 1), (1, 1, 0), (1, 0, 1), (1, 1, 1)])
pots = np.array([[0], [0], [0], [0], [0], [0], [0], [0]])
params = {'gdX': 0.2, 'gdY': 0.2, 'gdZ': 0.2, 'n_src': 10}
k = KCSD(elec_pos, pots, params)
k.estimate_pots()
k.estimate_csd()
for pot in k.solver.estimated_pots.flatten():
self.assertAlmostEqual(pot, 0.0, places=5)
for csd in k.solver.estimated_csd.flatten():
self.assertAlmostEqual(csd, 0.0, places=5)
def test_KCSD_1D_duplicated_electrode(self):
"""if two electrodes are at the same spot, it should raise exception"""
with self.assertRaises(Exception):
k = KCSD(elec_pos=np.array([[0], [0]]), sampled_pots=[[0], [0]])
def test_KCSD_1D_incorrect_electrode_number(self):
"""if there are more electrodes than pots, it should raise exception"""
with self.assertRaises(Exception):
k = KCSD(elec_pos=np.array([[0], [1], [2]]),
sampled_pots=[[0], [1]])
