Summary of Enhancing Compositional Generalization Via Compositional Feature Alignment, by Haoxiang Wang et al.
Enhancing Compositional Generalization via Compositional Feature Alignment
by Haoxiang Wang, Haozhe Si, Huajie Shao, Han Zhao
First submitted to arxiv on: 5 Feb 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG); Machine Learning (stat.ML)
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 tackles the challenge of training machine learning models to generalize well across multiple domains and classes. The authors introduce a new benchmark called CG-Bench, which evaluates the ability of models to learn from unseen domain-class combinations. They find that popular pre-trained models like CLIP and DINOv2 struggle with this task and propose a solution called Compositional Feature Alignment (CFA). CFA is a simple two-stage finetuning technique that learns to align features in a way that’s compositional, allowing the model to generalize better across unseen combinations. The authors show through experiments on CG-Bench that CFA outperforms other finetuning techniques in compositional generalization. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how machine learning models can learn from new situations they haven’t seen before. Right now, many models are only good at recognizing things they’ve been trained on, but this isn’t very useful for real-life applications. The authors created a special test to see if models can adapt to new situations and found that most pre-trained models don’t do well. They came up with a way called Compositional Feature Alignment (CFA) to help models learn in a way that lets them generalize better. This means they can recognize things even when the pictures are taken from different angles or have different lighting. |
Keywords
* Artificial intelligence * Alignment * Generalization * Machine learning
