Summary of Deep Active Learning with Manifold-preserving Trajectory Sampling, by Yingrui Ji et al.
Deep Active Learning with Manifold-preserving Trajectory Sampling
by Yingrui Ji, Vijaya Sindhoori Kaza, Nishanth Artham, Tianyang Wang
First submitted to arxiv on: 21 Oct 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 abstract discusses active learning (AL), a technique for optimizing the selection of unlabeled data for annotation to enhance model performance while minimizing labeling effort. The existing deep AL methods are biased towards labeled data, which is a significant issue in various types of data, including vision and non-vision data. To address this bias, the authors propose Manifold-Preserving Trajectory Sampling (MPTS), a novel method that enforces the feature space learned from labeled data to represent a more accurate manifold. This approach can be implemented by performing distribution mapping with Maximum Mean Discrepancies (MMD). The proposed method is evaluated on various benchmark datasets, demonstrating its superiority. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper talks about finding the best way to pick which data should get labeled so that machine learning models learn better while using less effort. The current methods for doing this have a problem because they favor the labeled data over the unlabeled data. This makes it hard to make good choices. To fix this, the authors came up with a new idea called Manifold-Preserving Trajectory Sampling (MPTS). It helps by making sure that the feature space learned from the labeled data is accurate and corrects the bias. They tested their method on different kinds of data and showed it works better. |
Keywords
» Artificial intelligence » Active learning » Machine learning
