Summary of Dino As a Von Mises-fisher Mixture Model, by Hariprasath Govindarajan et al.
DINO as a von Mises-Fisher mixture model
by Hariprasath Govindarajan, Per Sidén, Jacob Roll, Fredrik Lindsten
First submitted to arxiv on: 17 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
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 proposes an extension to self-distillation methods, specifically applying it to the popular DINO method, which uses Siamese networks for self-supervised pre-training. The authors interpret DINO and its derivatives as a mixture model of von Mises-Fisher components, allowing them to propose DINO-vMF, a modified version that adds normalization constants when computing cluster assignment probabilities. This modification improves image representations and outperforms the original DINO on various downstream tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper takes popular self-supervised pre-training methods like DINO and shows how they can be interpreted as a mixture model of von Mises-Fisher components. By adding some simple changes, the authors create a new version called DINO-vMF that performs better than the original DINO on many tasks. This is important because it means we can get even better results from our models just by making these small tweaks. |
Keywords
» Artificial intelligence » Distillation » Mixture model » Self supervised
