Summary of Robust Training Of Neural Networks at Arbitrary Precision and Sparsity, by Chengxi Ye et al.
Robust Training of Neural Networks at Arbitrary Precision and Sparsity
by Chengxi Ye, Grace Chu, Yanfeng Liu, Yichi Zhang, Lukasz Lew, Andrew Howard
First submitted to arxiv on: 14 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Numerical Analysis (math.NA)
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 solution stabilizes deep neural network training in ultra-low precision and sparse regimes by formulating quantization and sparsification as perturbations during training. A denoising affine transform is used to ensure robustness, and a piecewise constant backbone model provides a performance lower bound. This approach enables training of existing models at arbitrarily low precision and sparsity levels using off-the-shelf recipes. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Training deep neural networks in ultra-low precision and sparse regimes can be challenging due to discontinuous operations in quantization and sparsification. A new solution, the denoising affine transform, is proposed to stabilize training under these conditions. This approach uses ridge regression to derive a perturbation-resilient method that can train existing models at any level of precision or sparsity. |
Keywords
» Artificial intelligence » Neural network » Precision » Quantization » Regression
