Summary of Non-linear Outlier Synthesis For Out-of-distribution Detection, by Lars Doorenbos et al.
Non-Linear Outlier Synthesis for Out-of-Distribution Detection
by Lars Doorenbos, Raphael Sznitman, Pablo Márquez-Neila
First submitted to arxiv on: 20 Nov 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); 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 This paper addresses the limitations of supervised classifiers in dealing with unexpected inputs, a problem known as out-of-distribution (OOD) detection. Researchers have made progress by training OOD detectors on synthetic outliers generated by large diffusion models. The authors present NCIS, an approach that enhances the quality of these synthetic outliers by operating directly in the diffusion’s model embedding space and modeling class-conditional manifolds with a conditional volume-preserving network. This allows for more expressive characterization of the training distribution. The paper achieves new state-of-the-art OOD detection results on standard ImageNet100 and CIFAR100 benchmarks, providing insights into key design choices. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making sure computers can tell when something they’re supposed to recognize isn’t actually there. This is a big problem because it makes it hard for computers to make good decisions. Researchers have been trying to solve this by training special computers that can detect when something is weird. The new approach, called NCIS, does even better by using a special way of looking at the data and making sure the computer understands what’s normal. It works really well and shows how important it is to get the right kind of data. |
Keywords
» Artificial intelligence » Diffusion » Embedding space » Supervised
