Summary of Memory-efficient Prompt Tuning For Incremental Histopathology Classification, by Yu Zhu et al.
Memory-Efficient Prompt Tuning for Incremental Histopathology Classification
by Yu Zhu, Kang Li, Lequan Yu, Pheng-Ann Heng
First submitted to arxiv on: 22 Jan 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 proposed prompt tuning framework aims to upgrade a model’s generalization potential by learning from sequentially delivered domains without requiring massive computation resources. The approach reuses existing model parameters and attaches lightweight trainable prompts for customized tuning in each domain. Domain-specific prompts are used to investigate distinctive characteristics, while a shared domain-invariant prompt explores common content embedding throughout time. A graph with existing prompts is constructed to guide the domain-invariant prompt’s exploration of overlapped latent embeddings among all domains, leading to more domain-generic representations. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A recent paper has made progress in classifying histopathology images. The researchers propose a new way to improve their model by learning from different types of images without needing too many computer resources. They reuse the old model’s parameters and add small, trainable pieces that help the model learn about each type of image. Each type of image gets its own special prompt, which helps the model understand what makes it unique. The prompts are stored in a special bank to prevent the model from forgetting what it learned earlier. The researchers also use a shared prompt that changes over time to help the model become better at generalizing. They tested their approach on two types of images and found that it works well and uses less memory than other methods. |
Keywords
* Artificial intelligence * Embedding * Generalization * Prompt
