Summary of Enhancing Split Computing and Early Exit Applications Through Predefined Sparsity, by Luigi Capogrosso et al.
Enhancing Split Computing and Early Exit Applications through Predefined Sparsity
by Luigi Capogrosso, Enrico Fraccaroli, Giulio Petrozziello, Francesco Setti, Samarjit Chakraborty, Franco Fummi, Marco Cristani
First submitted to arxiv on: 16 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV); Distributed, Parallel, and Cluster Computing (cs.DC)
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 novel approach combines predefined sparsity with Split Computing (SC) and Early Exit (EE) to address the challenge of deploying Deep Neural Networks (DNNs) on resource-constrained edge devices. The SC splits a DNN, deploying part on an edge device and the rest on a remote server. EE allows early termination if the answer is good enough, reducing computational, storage, and energy burdens during training and inference phases. This paper studies predefined sparsity in SC and EE paradigms, demonstrating significant reductions in storage and computational complexity without compromising performance. Experimental results show exceeding 4x reductions in storage and computation. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper introduces a new way to use Deep Neural Networks (DNNs) on devices that don’t have much power or memory. The approach combines two ideas: splitting the DNN into parts, with some parts running on the device and others on a server; and stopping when the answer is good enough. This makes it possible to reduce the amount of work the device has to do, using less energy and storage space. The results show that this approach can make big savings without affecting how well the DNN works. |
Keywords
» Artificial intelligence » Inference
