Summary of Diversified Batch Selection For Training Acceleration, by Feng Hong et al.
Diversified Batch Selection for Training Acceleration
by Feng Hong, Yueming Lyu, Jiangchao Yao, Ya Zhang, Ivor W. Tsang, Yanfeng Wang
First submitted to arxiv on: 7 Jun 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 This paper presents Diversified Batch Selection (DivBS), a novel approach to online batch selection that efficiently selects diverse and representative samples without relying on reference models. The method addresses the limitations of previous methods by introducing a novel selection objective that measures group-wise orthogonalized representativeness, combating redundancy issues. Experimental results across various tasks demonstrate the significant superiority of DivBS in terms of performance-speedup trade-off. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about finding ways to make machine learning models more efficient. Right now, training these models takes a lot of time and resources because they need to look at huge amounts of data. Researchers have been trying to find ways to speed things up by selecting the most important parts of the data to look at. The problem is that some methods require a “perfect” model to work, which isn’t always available. Other methods try to do everything themselves and end up looking at too much of the same thing. This paper proposes a new way called Diversified Batch Selection (DivBS) that doesn’t need a perfect model and can efficiently pick out the most important parts of the data. The results show that this approach is better than others in terms of balancing performance and speed. |
Keywords
* Artificial intelligence * Machine learning
