Summary of Exploring Learngene Via Stage-wise Weight Sharing For Initializing Variable-sized Models, by Shi-yu Xia et al.
Exploring Learngene via Stage-wise Weight Sharing for Initializing Variable-sized Models
by Shi-Yu Xia, Wenxuan Zhu, Xu Yang, Xin Geng
First submitted to arxiv on: 25 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); 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 A machine learning framework called Learngene is introduced to build variable-sized models adapting for diverse resource constraints. The paper focuses on the importance of guidance for expanding well-trained learngene layers, which is crucial for initializing models at varying scales. To address this challenge, a simple yet effective approach called SWS (Stage-wise Weight Sharing) is designed. This method leverages an auxiliary model with shared layer weights across multiple stages to train and expand learngene layers. Extensive experiments on ImageNet-1K demonstrate that SWS achieves consistent better performance compared to many models trained from scratch, while reducing training costs by around 6.6x. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper introduces a new way to build variable-sized models that work well with limited resources. It’s like building a Lego tower – you start with a small base and then add more blocks as needed. The authors created a special type of layer called Learngene, which is trained to be useful for initializing different-sized models. They also developed an approach called SWS, which helps the model learn how to expand and adapt to new situations. The results show that this method can perform better than other approaches while using fewer resources. |
Keywords
» Artificial intelligence » Machine learning
