Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

orthonormalize subblocks when multiple species are present #18

Merged
merged 2 commits into from
Dec 5, 2023
Merged

orthonormalize subblocks when multiple species are present #18

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a5f2f9a
orthonormalize subblocks when multiple species are present
arthur-lin1027 Aug 15, 2023
0e138c6
added comments explaining orthonormalizing each subblock per neighbor…
arthur-lin1027 Aug 16, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
40 changes: 25 additions & 15 deletions anisoap/representations/radial_basis.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -202,20 +202,30 @@ def orthonormalize_basis(self, features: TensorMap):
)
return features
for label, block in features.items():
l = label["angular_channel"]
n_arr = block.properties["n"].flatten()
l_2n_arr = l + 2 * n_arr
# normalize all the GTOs by the appropriate prefactor first, since the overlap matrix is in terms of
# normalized GTOs
prefactor_arr = gto_prefactor(
l_2n_arr, self.hypers["radial_gaussian_width"]
)
block.values[:, :, :] = block.values[:, :, :] * prefactor_arr

gto_overlap_matrix_slice = self.overlap_matrix[l_2n_arr, :][:, l_2n_arr]
orthonormalization_matrix = inverse_matrix_sqrt(gto_overlap_matrix_slice)
block.values[:, :, :] = np.einsum(
"ijk,kl->ijl", block.values, orthonormalization_matrix
)
# Each block's `properties` dimension contains radial channels for each neighbor species
# Hence we have to iterate through each neighbor species and orthonormalize the block in subblocks
# Each subblock is indexed using the neighbor_mask boolean array.
neighbors = np.unique(block.properties["neighbor_species"])
for neighbor in neighbors:
l = label["angular_channel"]
neighbor_mask = block.properties["neighbor_species"] == neighbor
n_arr = block.properties["n"][neighbor_mask].flatten()
l_2n_arr = l + 2 * n_arr
# normalize all the GTOs by the appropriate prefactor first, since the overlap matrix is in terms of
# normalized GTOs
prefactor_arr = gto_prefactor(
l_2n_arr, self.hypers["radial_gaussian_width"]
)
block.values[:, :, neighbor_mask] *= prefactor_arr

gto_overlap_matrix_slice = self.overlap_matrix[l_2n_arr, :][:, l_2n_arr]
orthonormalization_matrix = inverse_matrix_sqrt(
gto_overlap_matrix_slice
)
block.values[:, :, neighbor_mask] = np.einsum(
"ijk,kl->ijl",
block.values[:, :, neighbor_mask],
orthonormalization_matrix,
)

return features
Loading