awslabs / awslabs/aws-cv-task2vec
Are task2vec embeddings valid when the network used doesn't perfectly fit the data?
Open
- Dominant language
- Python
- Stars
- 126
- Forks
- 31
- PR merge metrics
- No merged PRs in 30d
Description
e.g. what happens if one uses a randomly initialized probe network?
I am assuming it's likely invalid (but requires experiments to confirm) because FIM stuff usually needs to be close to an (global?) optimum.
Contributor guide
Assessment
This issue has not been assessed yet.