Summary of Forte : Finding Outliers with Representation Typicality Estimation, by Debargha Ganguly et al.
Forte : Finding Outliers with Representation Typicality Estimation
by Debargha Ganguly, Warren Morningstar, Andrew Yu, Vipin Chaudhary
First submitted to arxiv on: 2 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Information Theory (cs.IT)
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 Generative models can now produce photorealistic synthetic data that is virtually indistinguishable from real data used to train it. Recent work on OOD detection raised doubts about the effectiveness of generative model likelihoods as optimal OOD detectors due to issues involving likelihood misestimation, entropy in the generative process, and typicality. Our approach leverages representation learning and informative summary statistics based on manifold estimation to address these issues. We introduce a novel method that outperforms other unsupervised approaches and achieves state-of-the-art performance on challenging benchmarks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Generative models can now make fake pictures that look almost real. But some researchers thought that using these models to detect when something is “out of the ordinary” didn’t work very well. They think this might be because the models focus too much on what things look like, rather than what they mean. We have a new way to do this that’s better than before and works really well on lots of different tests. |
Keywords
» Artificial intelligence » Generative model » Likelihood » Representation learning » Synthetic data » Unsupervised
