def test_krr_bob():
test_dir = os.path.dirname(os.path.realpath(__file__))
# Parse file containing PBE0/def2-TZVP heats of formation and xyz filenames
data = get_energies(test_dir + "/data/hof_qm7.txt")
# Generate a list of qml.data.Compound() objects
mols = []
for xyz_file in sorted(data.keys()):
# Initialize the qml.data.Compound() objects
mol = qml.data.Compound(xyz=test_dir + "/qm7/" + xyz_file)
# Associate a property (heat of formation) with the object
mol.properties = data[xyz_file]
# This is a Molecular Coulomb matrix sorted by row norm
# mol.generate_eigenvalue_coulomb_matrix()
mol.generate_coulomb_matrix()
mols.append(mol)
# Shuffle molecules
np.random.seed(666)
np.random.shuffle(mols)
# Make training and test sets
n_test = 2000
n_train = 4000
training = mols[:n_train]
test = mols[-n_test:]
# List of representations
X = np.array([mol.representation for mol in training])
Xs = np.array([mol.representation for mol in test])
# List of properties
Y = np.array([mol.properties for mol in training])
Ys = np.array([mol.properties for mol in test])
# Set hyper-parameters
sigma = 4000.40
llambda = 1e-11
# Generate training Kernel
K = laplacian_kernel(X, X, sigma)
# Solve alpha
K[np.diag_indices_from(K)] += llambda
alpha = cho_solve(K, Y)
# Calculate prediction kernel
Ks = laplacian_kernel(X, Xs, sigma)
Yss = np.dot(Ks.transpose(), alpha)
mae = np.mean(np.abs(Ys - Yss))
print(mae)
def test_arad():
test_dir = os.path.dirname(os.path.realpath(__file__))
# Parse file containing PBE0/def2-TZVP heats of formation and xyz filenames
data = get_energies(test_dir + "/data/hof_qm7.txt")
# Generate a list of qml.data.Compound() objects
mols = []
for xyz_file in sorted(data.keys())[:10]:
# Initialize the qml.data.Compound() objects
mol = qml.data.Compound(xyz=test_dir + "/qm7/" + xyz_file)
# Associate a property (heat of formation) with the object
mol.properties = data[xyz_file]
# This is a Molecular Coulomb matrix sorted by row norm
mol.representation = generate_arad_representation(
mol.coordinates, mol.nuclear_charges)
mols.append(mol)
sigmas = [25.0]
X1 = np.array([mol.representation for mol in mols])
K_local_asymm = get_local_kernels_arad(X1, X1, sigmas)
K_local_symm = get_local_symmetric_kernels_arad(X1, sigmas)
assert np.allclose(K_local_symm,
K_local_asymm), "Symmetry error in local kernels"
assert np.invert(np.all(np.isnan(
K_local_asymm))), "ERROR: ARAD local symmetric kernel contains NaN"
K_global_asymm = get_global_kernels_arad(X1, X1, sigmas)
K_global_symm = get_global_symmetric_kernels_arad(X1, sigmas)
assert np.allclose(K_global_symm,
K_global_asymm), "Symmetry error in global kernels"
assert np.invert(np.all(np.isnan(
K_global_asymm))), "ERROR: ARAD global symmetric kernel contains NaN"
molid = 5
X1 = generate_arad_representation(mols[molid].coordinates,
mols[molid].nuclear_charges,
size=mols[molid].natoms)
XA = X1[:mols[molid].natoms]
K_atomic_asymm = get_atomic_kernels_arad(XA, XA, sigmas)
K_atomic_symm = get_atomic_symmetric_kernels_arad(XA, sigmas)
assert np.allclose(K_atomic_symm,
K_atomic_asymm), "Symmetry error in atomic kernels"
assert np.invert(np.all(np.isnan(
K_atomic_asymm))), "ERROR: ARAD atomic symmetric kernel contains NaN"
K_atomic_asymm = get_atomic_kernels_arad(XA, XA, sigmas)
def test_nn_bob():
test_dir = os.path.dirname(os.path.realpath(__file__))
# Parse file containing PBE0/def2-TZVP heats of formation and xyz filenames
data = get_energies(test_dir + "/data/hof_qm7.txt")
# Generate a list of qml.data.Compound() objects
mols = []
numbers = dict()
for xyz_file in sorted(data.keys()):
# Initialize the qml.data.Compound() objects
mol = qml.data.Compound(xyz=test_dir + "/qm7/" + xyz_file)
# Associate a property (heat of formation) with the object
mol.properties = data[xyz_file]
# This is a Molecular Coulomb matrix sorted by row norm
# mol.generate_eigenvalue_coulomb_matrix()
mol.generate_coulomb_matrix()
# mol.generate_bob()
# print(mol.representation)
mols.append(mol)
es = np.array([mol.properties for mol in mols])
fc, fa = get_mean_atomic_contribution(mols, es)
for i in range(len(mols)):
mols[i].properties = fc[i]
# Shuffle molecules
np.random.seed(666)
np.random.shuffle(mols)
# Make training and test sets
n_test = 2000
n_train = 4000
training = mols[:n_train]
test = mols[-n_test:]
# List of representations
X = np.array([mol.representation for mol in training])
Xs = np.array([mol.representation for mol in test])
# List of properties
Y = np.array([mol.properties for mol in training])
Ys = np.array([mol.properties for mol in test])
print(X.shape)
print(Y.shape)
dtype = torch.float
device = torch.device("cuda:0")
N, D_in, H, H2 = n_train, X.shape[1], 50, 10
x = torch.from_numpy(X).to(device, torch.float)
y = torch.from_numpy(Y.reshape((N, 1))).to(device, torch.float)
xs = torch.from_numpy(Xs).to(device, torch.float)
ys = torch.from_numpy(Ys.reshape((n_test, 1))).to(device, torch.float)
model = NeuralNet(X.shape[1], H, H2).to(device)
# Loss and optimizer
# criterion = nn.L1Loss()
criterion = nn.MSELoss()
optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)
for epoch in range(100000):
yt = model(x)
if (epoch % 1000 == 0):
loss = (y - yt).pow(2).sum()
rmsd = torch.sqrt(loss / n_train) * 627.51
mae = (y - yt).abs().sum() * 627.51 / n_train
yss = model.forward(xs)
rmsd_s = torch.sqrt((yss - ys).pow(2).sum() / n_test) * 627.51
mae_s = (yss - ys).abs().sum() * 627.51 / n_test
print(epoch, mae, rmsd, mae_s, rmsd_s)
loss = criterion(yt, y)
optimizer.zero_grad()
loss.backward()
optimizer.step()
# for epoch in range(100000):
# if (epoch % 1000 == 0):
# yt = model(x)
#
# loss =(y - yt).pow(2).sum()
# rmsd = torch.sqrt(loss/n_train) * 627.51
# mae = (y - yt).abs().sum() * 627.51 / n_train
# yss = model.forward(xs)
#
# rmsd_s = torch.sqrt((yss - ys).pow(2).sum()/n_test) * 627.51
# mae_s = (yss - ys).abs().sum() * 627.51 / n_test
# print(epoch, mae, rmsd, mae_s, rmsd_s)
#
# def closure():
# yt = model(x)
# return (y - yt).pow(2).sum()
# optimizer2.zero_grad()
# loss.backward(retain_graph=True)
# optimizer2.step(closure)
print(yt[:10] * 627.51, y[:10] * 627.51)
def test_nn_bob():
test_dir = os.path.dirname(os.path.realpath(__file__))
# Parse file containing PBE0/def2-TZVP heats of formation and xyz filenames
data = get_energies(test_dir + "/data/hof_qm7.txt")
# Generate a list of qml.data.Compound() objects
mols = []
numbers = dict()
for xyz_file in sorted(data.keys()):
# Initialize the qml.data.Compound() objects
mol = qml.data.Compound(xyz=test_dir + "/qm7/" + xyz_file)
# Associate a property (heat of formation) with the object
mol.properties = data[xyz_file]
# This is a Molecular Coulomb matrix sorted by row norm
mol.generate_eigenvalue_coulomb_matrix()
# print(mol.representation)
mols.append(mol)
# Shuffle molecules
# np.random.seed(666)
np.random.shuffle(mols)
# Make training and test sets
n_test = 1000
n_train = 1000
training = mols[:n_train]
test = mols[-n_test:]
# List of representations
X = np.array([mol.representation for mol in training])
Xs = np.array([mol.representation for mol in test])
# List of properties
Y = np.array([mol.properties for mol in training])
Ys = np.array([mol.properties for mol in test])
print(X.shape)
print(Y.shape)
# exit()
import torch
dtype = torch.float
# device = torch.device("cpu")
device = torch.device("cuda:0") # Uncomment this to run on GPU
N, D_in, H, D_out = n_train, X.shape[1], 64, 1
x = torch.from_numpy(X).to(device, torch.float)
y = torch.from_numpy(Y.reshape((N, 1))).to(device, torch.float)
xs = torch.from_numpy(Xs).to(device, torch.float)
ys = torch.from_numpy(Ys.reshape((n_test, 1))).to(device, torch.float)
# print(x.shape)
# print(y.shape)
# print(x)
# print(X)
# # Randomly initialize weights
# w1 = torch.randn(D_in, H, device=device, dtype=dtype)
# w2 = torch.randn(H, D_out, device=device, dtype=dtype)
# print(w1.shape)
# print(w2.shape)
# N, D_in, H, D_out = 64, 1000, 100, 10
# # Create random input and output data
# x = torch.randn(N, D_in, device=device, dtype=dtype)
# y = torch.randn(N, D_out, device=device, dtype=dtype)
# Randomly initialize weights
w1 = torch.randn(D_in, H, device=device, dtype=dtype)
w2 = torch.randn(H, D_out, device=device, dtype=dtype)
learning_rate = 1e-8
for t in range(5000):
# Forward pass: compute predicted y
h = x.mm(w1)
h_relu = h.clamp(min=0)
y_pred = h_relu.mm(w2)
# Compute and print loss
loss = (y_pred - y).pow(2).sum().item()
hval = xs.mm(w1)
hval_relu = hval.clamp(min=0)
yval_pred = hval_relu.mm(w2)
loss2 = (yval_pred - ys).pow(2).sum().item()
print(t, loss / n_train * 627.51, loss2 / n_test * 627.51)
# Backprop to compute gradients of w1 and w2 with respect to loss
grad_y_pred = 2.0 * (y_pred - y)
grad_w2 = h_relu.t().mm(grad_y_pred)
grad_h_relu = grad_y_pred.mm(w2.t())
grad_h = grad_h_relu.clone()
grad_h[h < 0] = 0
grad_w1 = x.t().mm(grad_h)
# Update weights using gradient descent
w1 -= learning_rate * grad_w1
w2 -= learning_rate * grad_w2
def test_krr_gaussian_local_cmat():
test_dir = os.path.dirname(os.path.realpath(__file__))
# Parse file containing PBE0/def2-TZVP heats of formation and xyz filenames
data = get_energies(test_dir + "/data/hof_qm7.txt")
# Generate a list of qml.data.Compound() objects"
mols = []
for xyz_file in sorted(data.keys())[:1000]:
# Initialize the qml.data.Compound() objects
mol = qml.data.Compound(xyz=test_dir + "/qm7/" + xyz_file)
# Associate a property (heat of formation) with the object
mol.properties = data[xyz_file]
# This is a Molecular Coulomb matrix sorted by row norm
mol.generate_atomic_coulomb_matrix(size=23, sorting="row-norm")
mols.append(mol)
# Shuffle molecules
np.random.seed(666)
np.random.shuffle(mols)
# Make training and test sets
n_test = 100
n_train = 200
training = mols[:n_train]
test = mols[-n_test:]
X = np.concatenate([mol.representation for mol in training])
Xs = np.concatenate([mol.representation for mol in test])
N = np.array([mol.natoms for mol in training])
Ns = np.array([mol.natoms for mol in test])
# List of properties
Y = np.array([mol.properties for mol in training])
Ys = np.array([mol.properties for mol in test])
# Set hyper-parameters
sigma = 724.0
llambda = 10**(-6.5)
K = get_local_kernels_gaussian(X, X, N, N, [sigma])[0]
assert np.allclose(K, K.T), "Error in local Gaussian kernel symmetry"
K_test = np.loadtxt(test_dir + "/data/K_local_gaussian.txt")
assert np.allclose(
K, K_test), "Error in local Gaussian kernel (vs. reference)"
K_test = get_atomic_kernels_gaussian(training, training, [sigma])[0]
assert np.allclose(K,
K_test), "Error in local Gaussian kernel (vs. wrapper)"
# Solve alpha
K[np.diag_indices_from(K)] += llambda
alpha = cho_solve(K, Y)
# Calculate prediction kernel
Ks = get_local_kernels_gaussian(Xs, X, Ns, N, [sigma])[0]
Ks_test = np.loadtxt(test_dir + "/data/Ks_local_gaussian.txt")
# Somtimes a few coulomb matrices differ because of parallel sorting and numerical error
# Allow up to 5 molecules to differ from the supplied reference.
differences_count = len(set(np.where(Ks - Ks_test > 1e-7)[0]))
assert differences_count < 5, "Error in local Laplacian kernel (vs. reference)"
# assert np.allclose(Ks, Ks_test), "Error in local Gaussian kernel (vs. reference)"
Ks_test = get_atomic_kernels_gaussian(test, training, [sigma])[0]
assert np.allclose(Ks,
Ks_test), "Error in local Gaussian kernel (vs. wrapper)"
Yss = np.dot(Ks, alpha)
mae = np.mean(np.abs(Ys - Yss))
print(mae)
assert abs(19.0 -
mae) < 1.0, "Error in local Gaussian kernel-ridge regression"
def test_krr_laplacian_local_cmat():
test_dir = os.path.dirname(os.path.realpath(__file__))
# Parse file containing PBE0/def2-TZVP heats of formation and xyz filenames
data = get_energies(test_dir + "/data/hof_qm7.txt")
# Generate a list of qml.data.Compound() objects"
mols = []
for xyz_file in sorted(data.keys())[:1000]:
# Initialize the qml.data.Compound() objects
mol = qml.data.Compound(xyz=test_dir + "/qm7/" + xyz_file)
# Associate a property (heat of formation) with the object
mol.properties = data[xyz_file]
# This is a Molecular Coulomb matrix sorted by row norm
mol.generate_atomic_coulomb_matrix(size=23, sorting="row-norm")
mols.append(mol)
# Shuffle molecules
np.random.seed(666)
np.random.shuffle(mols)
# Make training and test sets
n_test = 100
n_train = 200
training = mols[:n_train]
test = mols[-n_test:]
X = np.concatenate([mol.representation for mol in training])
Xs = np.concatenate([mol.representation for mol in test])
N = np.array([mol.natoms for mol in training])
Ns = np.array([mol.natoms for mol in test])
# List of properties
Y = np.array([mol.properties for mol in training])
Ys = np.array([mol.properties for mol in test])
# Set hyper-parameters
sigma = 10**(3.6)
llambda = 10**(-12.0)
K = get_local_kernels_laplacian(X, X, N, N, [sigma])[0]
assert np.allclose(K, K.T), "Error in local Laplacian kernel symmetry"
# Test below will sometimes fail, since sorting occasionally differs due close row-norms
# K_test = np.loadtxt(test_dir + "/data/K_local_laplacian.txt")
# assert np.allclose(K, K_test), "Error in local Laplacian kernel (vs. reference)"
# Solve alpha
K[np.diag_indices_from(K)] += llambda
alpha = cho_solve(K, Y)
# Calculate prediction kernel
Ks = get_local_kernels_laplacian(Xs, X, Ns, N, [sigma])[0]
# Test below will sometimes fail, since sorting occasionally differs due close row-norms
# Ks_test = np.loadtxt(test_dir + "/data/Ks_local_laplacian.txt")
# assert np.allclose(Ks, Ks_test), "Error in local Laplacian kernel (vs. reference)"
Yss = np.dot(Ks, alpha)
mae = np.mean(np.abs(Ys - Yss))
assert abs(8.7 -
mae) < 1.0, "Error in local Laplacian kernel-ridge regression"