Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Hello,
I think the current implementation of CoxLoss is not accurate, below I discuss the two issues I identified and their solutions:
loss_cox = -torch.mean((theta - torch.log(torch.sum(exp_theta*R_mat, dim=1))) * censor)
this part
(theta - torch.log(torch.sum(exp_theta*R_mat, dim=1)))
returns a 1d tensor, ifcensor
is a 2d tensor (which is the case most of the time, as it is not squeezed in the functionCoxLoss
), then pytorch will have to broadcast, and this multiplication will return a 2d matrix.Toy example:
The fix to this is to ensure
censor
is a 1d tensor, as I propose here.Thanks