Summary of Transfer-learning-based Autotuning Using Gaussian Copula, by Thomas Randall (1) et al.
Transfer-Learning-Based Autotuning Using Gaussian Copula
by Thomas Randall, Jaehoon Koo, Brice Videau, Michael Kruse, Xingfu Wu, Paul Hovland, Mary Hall, Rong Ge, Prasanna Balaprakash
First submitted to arxiv on: 9 Jan 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 In this paper, researchers propose a new approach to autotuning for high-performance computing (HPC) systems. Autotuning is an empirical method that adjusts parameters to optimize performance, but it can be computationally expensive. The authors introduce a transfer learning-based approach called Gaussian copula (GC) that leverages data from prior tuning to generate high-performing configurations for new tasks. This allows for efficient few-shot tuning and provides a probabilistic estimation of the few-shot budget. The GC model achieves up to 64.37% of peak few-shot performance in its first evaluation, outperforming state-of-the-art techniques. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about using computers to solve big problems. Right now, people are building super-powerful computers that can do lots of things at once. But it’s hard to make these computers work efficiently. One way to fix this is by using a special technique called autotuning. Autotuning tries different settings and sees what works best. The problem is that this method takes a long time and uses too many resources. A new approach uses something called transfer learning, which helps the computer learn from previous experiences. This makes it possible to find good settings quickly and use them for new tasks. The results show that this new approach can be up to 33 times faster than the old way! |
Keywords
* Artificial intelligence * Few shot * Transfer learning
