Summary of Zero-shot Nas Via the Suppression Of Local Entropy Decrease, by Ning Wu et al.
Zero-Shot NAS via the Suppression of Local Entropy Decrease
by Ning Wu, Han Huang, Yueting Xu, Zhifeng Hao
First submitted to arxiv on: 9 Nov 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV); Neural and Evolutionary Computing (cs.NE)
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 novel zero-shot neural architecture search (NAS) approach accelerates evaluation by leveraging a data-free and running-free proxy, dubbed Suppression of Local Entropy Decrease (SED). By utilizing architectural topologies to quantify SED, the proposed method outperforms state-of-the-art proxies on five benchmarks, reducing computation time by three orders of magnitude. The SED-based NAS selects optimal architectures with higher accuracy and fewer parameters in just one second, demonstrating its potential for efficient architecture selection. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A team of researchers has developed a new way to quickly evaluate and select the best neural network architecture without needing large amounts of data or computation time. They call this method “Suppression of Local Entropy Decrease” or SED for short. By using special architectural designs, they can predict how well a network will perform without actually training it on any data. This approach is much faster and more efficient than current methods, which require running the networks on lots of data and doing many calculations. The researchers show that their method works well on different types of tasks and selects the best architectures in just a second. |
Keywords
» Artificial intelligence » Neural network » Zero shot
