Summary of Semi-supervised Variational Adversarial Active Learning Via Learning to Rank and Agreement-based Pseudo Labeling, by Zongyao Lyu et al.
Semi-Supervised Variational Adversarial Active Learning via Learning to Rank and Agreement-Based Pseudo Labeling
by Zongyao Lyu, William J. Beksi
First submitted to arxiv on: 23 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: 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 This paper proposes novel techniques to improve active learning, specifically addressing shortcomings in variational adversarial active learning (VAAL). The authors introduce a pseudo-labeling algorithm that leverages abundant unlabeled data for semi-supervised training, allowing models to explore richer data spaces. They also develop a ranking-based loss prediction module that embeds predicted relative rankings into the latent space of a variational autoencoder, enabling sample selection through adversarial training. This approach outperforms the state-of-the-art on various image classification and segmentation benchmark datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps machines learn from data better by using unlabeled samples in a smarter way. Right now, some methods only use these samples to pick which ones to label next, but this paper shows how to also use them to train the model itself. They came up with two new techniques: one that labels the data in a semi-supervised way and another that ranks the samples based on their similarity to labeled data. This combination lets the model explore more of the data space and makes it better at classifying images and segmenting objects. |
Keywords
» Artificial intelligence » Active learning » Image classification » Latent space » Semi supervised » Variational autoencoder
