Summary of Going Beyond Conventional Ood Detection, by Sudarshan Regmi
Going Beyond Conventional OOD Detection
by Sudarshan Regmi
First submitted to arxiv on: 16 Nov 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper presents a unified approach to detecting out-of-distribution (OOD) samples, particularly in challenging scenarios where spurious correlations and fine-grained classification are involved. The proposed method, ASCOOD, synthesizes virtual outliers from in-distribution data by approximating the destruction of invariant features using pixel attribution. This eliminates the need for external OOD datasets. ASCOOD simultaneously incentivizes ID classification and predictive uncertainty towards virtual outliers, leveraging standardized feature representation. The approach effectively mitigates spurious correlations and captures fine-grained attributes. Experiments across seven datasets demonstrate ASCOOD’s merit in spurious, fine-grained, and conventional settings. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary ASCOOD is a new way to detect when something doesn’t belong. This happens often with deep learning models that can mistake things they shouldn’t know about as familiar. The problem gets worse if the training data has misleading connections. ASCOOD makes it easier to find these outliers by creating fake ones from the normal data. It also makes sure the model is good at both correctly identifying things and being unsure when it’s not sure. This helps the model focus on important details. ASCOOD works well in different situations, including tricky cases where things are very similar. |
Keywords
* Artificial intelligence * Classification * Deep learning
