Summary of You Can’t Handle the (dirty) Truth: Data-centric Insights Improve Pseudo-labeling, by Nabeel Seedat et al.
You can’t handle the (dirty) truth: Data-centric insights improve pseudo-labeling
by Nabeel Seedat, Nicolas Huynh, Fergus Imrie, Mihaela van der Schaar
First submitted to arxiv on: 19 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- 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 This paper proposes a novel approach to semi-supervised learning called Pseudo-Labeling, which leverages unlabeled data when labeled samples are scarce. The method relies on generating and selecting pseudo-labels, but existing approaches assume that the labeled data is perfect. However, this assumption can be violated in reality due to issues like mislabeling or ambiguity. To address this overlooked aspect, the authors introduce a framework called DIPS (Data Inspection and Pseudo-Label Selection) that characterizes and selects useful labeled and pseudo-labeled samples based on learning dynamics analysis. The method is demonstrated across various real-world datasets, improving data efficiency and reducing performance distinctions between different pseudo-labelers. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Pseudo-labeling is a way to use extra data when we don’t have enough labeled information. It works by pretending the unlabeled data has labels, but this can be tricky because sometimes the labeled data might not be perfect. The authors of this paper want to make sure that the pseudo-labeling method works well even if the labeled data is messy. They came up with a new way to look at the data and pick out the good parts, which they call DIPS. This helps make the pseudo-labeling method better and more efficient. |
Keywords
» Artificial intelligence » Semi supervised
