Summary of Fr-nas: Forward-and-reverse Graph Predictor For Efficient Neural Architecture Search, by Haoming Zhang and Ran Cheng
FR-NAS: Forward-and-Reverse Graph Predictor for Efficient Neural Architecture Search
by Haoming Zhang, Ran Cheng
First submitted to arxiv on: 24 Apr 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 proposed neural architecture search (NAS) method leverages Graph Neural Networks (GNNs) to predict the performance of deep neural networks without exhaustive training. The approach combines conventional and inverse graph views to generate vector representations of neural architectures, which are then optimized using a customized training loss function. Experimental results on benchmark datasets NAS-Bench-101, NAS-Bench-201, and DARTS search space demonstrate a significant improvement in prediction accuracy, with a 3%–16% increase in Kendall-tau correlation. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Neural Architecture Search (NAS) helps find the best deep neural networks for specific tasks. To make this process faster, researchers use something called performance predictors that can estimate how well an architecture will work without having to train it all the way. This is similar to using a map to predict where you’ll end up, rather than actually driving there first. The problem is that these predictors need training data, which can be scarce. To solve this issue, scientists created a new predictor that combines two ways of looking at neural architectures: from the front (conventional) and from the back (inverse). They also designed a special way to train this predictor so it uses its knowledge efficiently. The team tested their method on various datasets and found that it worked better than other predictors, with an improvement in prediction accuracy. |
Keywords
» Artificial intelligence » Loss function
