Skip to content

Fix redundancy in OOD Detection description #418

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

Merged
sergiopaniego merged 1 commit into from
Jan 14, 2026
Merged

Fix redundancy in OOD Detection description #418

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 1 addition & 1 deletion chapters/en/unit11/1.mdx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -98,7 +98,7 @@ We’ve gone a long way! We now have a decent idea of zero-shot learning, its hi

- **Domain Adaptation (DA)**: This should be familiar by now. We can think of Zero-shot Learning as an extreme case of Domain Adaptation (greedy, greedy), dealing with the problem of learning to recognize unseen classes with no data at all. Domain Adaptation focuses on bridging the gap between two related domains (datasets) with different distributions and it requires labeled data.
- **Open Set Recognition (OSR)**: Think of it as a Boolean Zero-shot Learning. Like ZSL, in the OSR problem, the model handles both seen and unseen classes. But unlike ZSL, the model only needs to classify, at test time, whether the instance belongs to the seen classes or not. That’s it, no fancy labels. Still, this is a significant challenge.
- **Out of Distribution (OOD) Detection**: Think of this problem as a continuous variant of open set recognition. Here we don’t want to detect any instances that weren’t included in the training process, but rather just instances that deviate significantly from the training data distribution distribution. By recognizing and handling unexpected data effectively, OOD detection can pave the way for more trustworthy and robust AI systems that can adapt to unpredictable environments.
- **Out of Distribution (OOD) Detection**: Think of this problem as a continuous variant of open set recognition. Here we don’t want to detect any instances that weren’t included in the training process, but rather just instances that deviate significantly from the training data distribution. By recognizing and handling unexpected data effectively, OOD detection can pave the way for more trustworthy and robust AI systems that can adapt to unpredictable environments.
- **Open Vocabulary Learning (OVL)**: This is what zero-shot learning aspires to be. Like, ZSL on steroids. Overall, OVL can be considered an extension of ZSL, encompassing the ability to learn from limited data for unseen classes while also extending to handle seen classes and potentially infinite new classes and tasks.

## References
Expand Down