Summary of Generalized Semi-supervised Learning Via Self-supervised Feature Adaptation, by Jiachen Liang et al.
Generalized Semi-Supervised Learning via Self-Supervised Feature Adaptation
by Jiachen Liang, Ruibing Hou, Hong Chang, Bingpeng Ma, Shiguang Shan, Xilin Chen
First submitted to arxiv on: 31 May 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 paper proposes a novel semi-supervised learning (SSL) setting where unlabeled data deviates from the feature distribution of labeled samples. Traditional SSL methods predict wrong pseudo-labels due to noise accumulation. The authors introduce Self-Supervised Feature Adaptation (SSFA), a framework that decouples pseudo-label prediction from model fitting to improve SSL performance. SSFA incorporates a self-supervised task to adapt the feature extractor to unlabeled data, generating high-quality pseudo-labels. The approach is applicable to various pseudo-label-based SSL learners and improves performance in labeled, unlabeled, and unseen distributions. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper finds a new way to do machine learning when we have some information that’s labeled and some that isn’t. Usually, this type of learning assumes the labeled and unlabeled data look similar, but often they don’t. This makes it hard for machines to learn from both types of data. The authors came up with an idea called Self-Supervised Feature Adaptation (SSFA) that helps machines learn better from a mix of labeled and unlabeled data. SSFA works by using some of the unlabeled data to adjust how the machine looks at the features in the data, making it better at predicting what’s in the unlabeled data. |
Keywords
» Artificial intelligence » Machine learning » Self supervised » Semi supervised
