Skip to content

detach inverse in the VAMPE score #274

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

Draft
DanielWicz wants to merge 2 commits into
base: main
Choose a base branch
from
Draft

detach inverse in the VAMPE score #274

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
795352d
detach invers in VAMPE score
DanielWicz May 11, 2023
0bb3215
Detach koopman from the graph, to prevent gradient update
DanielWicz May 11, 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
8 changes: 5 additions & 3 deletions deeptime/decomposition/deep/_vampnet.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -213,9 +213,11 @@ def vamp_score(data: "torch.Tensor", data_lagged: "torch.Tensor", method='VAMP2'
out = torch.pow(torch.norm(koopman, p='fro'), 2)
elif method == 'VAMPE':
c00, c0t, ctt = covariances(data, data_lagged, remove_mean=True)
c00_sqrt_inv = sym_inverse(c00, epsilon=epsilon, return_sqrt=True, mode=mode)
ctt_sqrt_inv = sym_inverse(ctt, epsilon=epsilon, return_sqrt=True, mode=mode)
koopman = multi_dot([c00_sqrt_inv, c0t, ctt_sqrt_inv]).t()
# in original paper of VAMPE, inv can be detached from gradient
c00_sqrt_inv = sym_inverse(c00, epsilon=epsilon, return_sqrt=True, mode=mode).detach()
ctt_sqrt_inv = sym_inverse(ctt, epsilon=epsilon, return_sqrt=True, mode=mode).detach()
# detach koopman, so that VAMPE is only depedent on the trace
koopman = multi_dot([c00_sqrt_inv, c0t, ctt_sqrt_inv]).t().detach()

u, s, v = torch.svd(koopman)
mask = s > epsilon
Expand Down