def test_ts(self):
from pele.utils.xyz import read_xyz
path = os.path.dirname(os.path.abspath(__file__))
xyz = read_xyz(open(path + "/lj18_ts.xyz", "r"))
x = xyz.coords.flatten()
vec = np.random.rand(x.size)
vec /= np.linalg.norm(vec)
self.eigpot = LowestEigPot(x,
self.pot,
orthogZeroEigs=self.system.
get_orthogonalize_to_zero_eigenvectors(),
first_order=self.first_order)
e1 = self.eigpot.getEnergy(vec)
e, g = self.eigpot.getEnergyGradient(vec)
gnum = self.eigpot.NumericalDerivative(vec, eps=1e-4)
gnum -= np.dot(gnum, vec)
self.assertLess(old_div(np.max(np.abs(g - gnum)), np.max(np.abs(g))),
1e-3)
self.assertAlmostEqual(e, e1, delta=e * 1e-4)
self.assertAlmostEqual(
1.,
np.dot(g, gnum) / np.linalg.norm(g) / np.linalg.norm(gnum), 3)
class TestEigPot(unittest.TestCase):
def setUp(self):
self.first_order = False
self.setUp1()
def setUp1(self):
np.random.seed(0)
natoms = 18
# s = LJCluster(natoms)
# nfrozen = 6
# reference_coords = s.get_random_configuration()
# self.system = LJClusterFrozen(13, range(nfrozen), reference_coords)
self.system = LJCluster(natoms)
self.x = self.system.get_random_configuration()
#ret = self.system.get_random_minimized_configuration(tol=100.)
#self.x = ret.coords
self.pot = self.system.get_potential()
self.eigpot = LowestEigPot(self.x, self.pot,
orthogZeroEigs=self.system.get_orthogonalize_to_zero_eigenvectors(),
first_order=self.first_order,
)
def test_gradient(self):
vec = np.random.rand(self.x.size)
vec /= np.linalg.norm(vec)
e1 = self.eigpot.getEnergy(vec)
e, g = self.eigpot.getEnergyGradient(vec)
gnum = self.eigpot.NumericalDerivative(vec, eps=1e-6)
gnum -= np.dot(gnum, vec)
self.assertLess(np.max(np.abs(g-gnum)) / np.max(np.abs(g)), 1e-2)
self.assertAlmostEqual(e, e1, delta=e * 1e-4)
self.assertAlmostEqual(1., np.dot(g, gnum) / np.linalg.norm(g) / np.linalg.norm(gnum), 3)
def test_ts(self):
from pele.utils.xyz import read_xyz
path = os.path.dirname(os.path.abspath(__file__))
xyz = read_xyz(open(path + "/lj18_ts.xyz", "r"))
x = xyz.coords.flatten()
vec = np.random.rand(x.size)
vec /= np.linalg.norm(vec)
self.eigpot = LowestEigPot(x, self.pot,
orthogZeroEigs=self.system.get_orthogonalize_to_zero_eigenvectors(),
first_order=self.first_order)
e1 = self.eigpot.getEnergy(vec)
e, g = self.eigpot.getEnergyGradient(vec)
gnum = self.eigpot.NumericalDerivative(vec, eps=1e-4)
gnum -= np.dot(gnum, vec)
self.assertLess(np.max(np.abs(g-gnum)) / np.max(np.abs(g)), 1e-3)
self.assertAlmostEqual(e, e1, delta=e * 1e-4)
self.assertAlmostEqual(1., np.dot(g, gnum) / np.linalg.norm(g) / np.linalg.norm(gnum), 3)
def setUp1(self):
np.random.seed(0)
natoms = 18
# s = LJCluster(natoms)
# nfrozen = 6
# reference_coords = s.get_random_configuration()
# self.system = LJClusterFrozen(13, range(nfrozen), reference_coords)
self.system = LJCluster(natoms)
self.x = self.system.get_random_configuration()
#ret = self.system.get_random_minimized_configuration(tol=100.)
#self.x = ret.coords
self.pot = self.system.get_potential()
self.eigpot = LowestEigPot(
self.x,
self.pot,
orthogZeroEigs=self.system.get_orthogonalize_to_zero_eigenvectors(
),
first_order=self.first_order,
)
def setUp1(self):
np.random.seed(0)
natoms = 18
# s = LJCluster(natoms)
# nfrozen = 6
# reference_coords = s.get_random_configuration()
# self.system = LJClusterFrozen(13, range(nfrozen), reference_coords)
self.system = LJCluster(natoms)
self.x = self.system.get_random_configuration()
#ret = self.system.get_random_minimized_configuration(tol=100.)
#self.x = ret.coords
self.pot = self.system.get_potential()
self.eigpot = LowestEigPot(self.x, self.pot,
orthogZeroEigs=self.system.get_orthogonalize_to_zero_eigenvectors(),
first_order=self.first_order,
)
def test_ts(self):
from pele.utils.xyz import read_xyz
path = os.path.dirname(os.path.abspath(__file__))
xyz = read_xyz(open(path + "/lj18_ts.xyz", "r"))
x = xyz.coords.flatten()
vec = np.random.rand(x.size)
vec /= np.linalg.norm(vec)
self.eigpot = LowestEigPot(x, self.pot,
orthogZeroEigs=self.system.get_orthogonalize_to_zero_eigenvectors(),
first_order=self.first_order)
e1 = self.eigpot.getEnergy(vec)
e, g = self.eigpot.getEnergyGradient(vec)
gnum = self.eigpot.NumericalDerivative(vec, eps=1e-4)
gnum -= np.dot(gnum, vec)
self.assertLess(np.max(np.abs(g-gnum)) / np.max(np.abs(g)), 1e-3)
self.assertAlmostEqual(e, e1, delta=e * 1e-4)
self.assertAlmostEqual(1., np.dot(g, gnum) / np.linalg.norm(g) / np.linalg.norm(gnum), 3)
class TestEigPot(unittest.TestCase):
def setUp(self):
self.first_order = False
self.setUp1()
def setUp1(self):
np.random.seed(0)
natoms = 18
# s = LJCluster(natoms)
# nfrozen = 6
# reference_coords = s.get_random_configuration()
# self.system = LJClusterFrozen(13, range(nfrozen), reference_coords)
self.system = LJCluster(natoms)
self.x = self.system.get_random_configuration()
#ret = self.system.get_random_minimized_configuration(tol=100.)
#self.x = ret.coords
self.pot = self.system.get_potential()
self.eigpot = LowestEigPot(
self.x,
self.pot,
orthogZeroEigs=self.system.get_orthogonalize_to_zero_eigenvectors(
),
first_order=self.first_order,
)
def test_gradient(self):
vec = np.random.rand(self.x.size)
vec /= np.linalg.norm(vec)
e1 = self.eigpot.getEnergy(vec)
e, g = self.eigpot.getEnergyGradient(vec)
gnum = self.eigpot.NumericalDerivative(vec, eps=1e-6)
gnum -= np.dot(gnum, vec)
self.assertLess(old_div(np.max(np.abs(g - gnum)), np.max(np.abs(g))),
1e-2)
self.assertAlmostEqual(e, e1, delta=e * 1e-4)
self.assertAlmostEqual(
1.,
np.dot(g, gnum) / np.linalg.norm(g) / np.linalg.norm(gnum), 3)
def test_ts(self):
from pele.utils.xyz import read_xyz
path = os.path.dirname(os.path.abspath(__file__))
xyz = read_xyz(open(path + "/lj18_ts.xyz", "r"))
x = xyz.coords.flatten()
vec = np.random.rand(x.size)
vec /= np.linalg.norm(vec)
self.eigpot = LowestEigPot(x,
self.pot,
orthogZeroEigs=self.system.
get_orthogonalize_to_zero_eigenvectors(),
first_order=self.first_order)
e1 = self.eigpot.getEnergy(vec)
e, g = self.eigpot.getEnergyGradient(vec)
gnum = self.eigpot.NumericalDerivative(vec, eps=1e-4)
gnum -= np.dot(gnum, vec)
self.assertLess(old_div(np.max(np.abs(g - gnum)), np.max(np.abs(g))),
1e-3)
self.assertAlmostEqual(e, e1, delta=e * 1e-4)
self.assertAlmostEqual(
1.,
np.dot(g, gnum) / np.linalg.norm(g) / np.linalg.norm(gnum), 3)