Summary of Stochastic Two Points Method For Deep Model Zeroth-order Optimization, by Yijiang Pang et al.
Stochastic Two Points Method for Deep Model Zeroth-order Optimization
by Yijiang Pang, Jiayu Zhou
First submitted to arxiv on: 2 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 paper introduces an efficient Stochastic Two-Point (S2P) approach, a zeroth-order method that updates large foundation models using only forward passes, without requiring backpropagation. The S2P approach offers a promising direction for tackling the challenge of building or fine-tuning such large models, which is often prohibitive due to hardware budget constraints. The paper presents theoretical convergence properties of S2P under general and relaxed smoothness assumptions, connecting it to popular types of zeroth-order methods like basic random search and stochastic three-point method. Empirical results show that the Variant of S2P (VS2P) outperforms or achieves competitive performance compared to standard methods across various model types and scales. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large foundation models are really good at many things, but they can be hard to make because we need powerful computers. This paper helps with that problem by introducing a new way to update these big models using only simple calculations, without needing special computer power. The new method is called Stochastic Two-Point (S2P) and it’s really good at making big models work well. |
Keywords
* Artificial intelligence * Backpropagation * Fine tuning
