Summary of Pseudo-labelling Meets Label Smoothing For Noisy Partial Label Learning, by Darshana Saravanan et al.
Pseudo-labelling meets Label Smoothing for Noisy Partial Label Learning
by Darshana Saravanan, Naresh Manwani, Vineet Gandhi
First submitted to arxiv on: 7 Feb 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 proposed work focuses on Noisy Partial Label Learning (NPLL), a weakly-supervised learning paradigm that relaxes the constraint of traditional Partial Label Learning (PLL) by allowing some partial labels to not contain the true label. The authors present a minimalistic framework that initially assigns pseudo-labels to images using a weighted nearest neighbour algorithm, and then train a deep neural network classifier with label smoothing. These features and predictions are used to refine and enhance the accuracy of pseudo-labels. Experimental results on seven datasets show state-of-the-art performance in all studied settings, including fine-grained classification and extreme noise scenarios. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary NPLL is a way to teach machines to learn from noisy or incomplete labels. Imagine you’re trying to teach a machine to recognize different animal species, but some of the training examples are wrong or missing information. This paper proposes a new method for NPLL that does better than previous methods in several ways. It uses a combination of algorithms and neural networks to refine its predictions and make more accurate decisions. The results show that this approach works well on a variety of datasets, including ones with extreme noise or incomplete labels. |
Keywords
* Artificial intelligence * Classification * Neural network * Supervised
