Summary of Rapid Deployment Of Dnns For Edge Computing Via Structured Pruning at Initialization, by Bailey J. Eccles et al.
Rapid Deployment of DNNs for Edge Computing via Structured Pruning at Initialization
by Bailey J. Eccles, Leon Wong, Blesson Varghese
First submitted to arxiv on: 22 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 paper explores a new approach called Pruning-at-Initialization (PaI) for compressing deep neural networks (DNNs) suitable for edge machine learning, which requires localized processing on devices. Traditional pruning methods compromise model performance or require substantial compute resources and time. The proposed PaI method uses structured pruning to mitigate these issues. The system, named Reconvene, rapidly generates pruned models suited for edge deployments by identifying and pruning convolution layers that are least sensitive to structured pruning. This approach achieves up to 16.21x size reduction and 2x speedup while maintaining the same accuracy as an unstructured PaI counterpart. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Edge machine learning is made possible by deep neural networks, but these models require a lot of computing power, memory, and energy to run on devices. To make them work better for edge machines, researchers have tried to compress the models using pruning techniques. However, existing methods either don’t improve performance or take too long to create the compressed models. This paper introduces a new way to prune deep neural networks called Pruning-at-Initialization (PaI). The PaI method uses structured pruning to make the models smaller and faster without losing their accuracy. The system, named Reconvene, is designed to quickly generate pruned models that are perfect for edge machines. |
Keywords
» Artificial intelligence » Machine learning » Pruning
