Summary of Disentangled Noisy Correspondence Learning, by Zhuohang Dang et al.
Disentangled Noisy Correspondence Learning
by Zhuohang Dang, Minnan Luo, Jihong Wang, Chengyou Jia, Haochen Han, Herun Wan, Guang Dai, Xiaojun Chang, Jingdong Wang
First submitted to arxiv on: 10 Aug 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
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 a novel framework called DisNCL (Disentanglement in Noisy Correspondence Learning) for cross-modal retrieval in noisy correspondence scenarios. The existing methods assume well-matched training data, which is impractical as real-world data involves imperfect alignments and noise. DisNCL introduces an information-theoretic approach to adaptively balance the extraction of modality-invariant information (MII) and modality-exclusive information (MEI) with certifiable optimal cross-modal disentanglement efficacy. This framework enhances similarity predictions in a modality-invariant subspace, boosting alleviation strategies for noisy correspondences. The paper also introduces soft matching targets to model noisy many-to-many relationships in multi-modal inputs. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Cross-modal retrieval is important for understanding connections between different types of data. However, this can be challenging when the data isn’t perfect or well-organized. Researchers have tried to find ways to make this process work better, but it’s still a problem. This paper presents a new way called DisNCL that uses information theory to help sort through the noise and find meaningful connections between different types of data. |
Keywords
» Artificial intelligence » Boosting » Multi modal
