Summary of Carbon Intensity-aware Adaptive Inference Of Dnns, by Jiwan Jung
Carbon Intensity-Aware Adaptive Inference of DNNs
by Jiwan Jung
First submitted to arxiv on: 23 Mar 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 proposes an innovative approach to make DNN inference more sustainable by adapting model size and accuracy to varying carbon intensity throughout the day. A heuristic algorithm is developed, which uses larger, high-accuracy models during low-intensity periods and smaller, lower-accuracy ones during high-intensity periods. Additionally, a new metric called carbon-emission efficiency is introduced to quantify the effectiveness of adaptive model selection in reducing carbon footprint. The evaluation shows that this approach can improve carbon emission efficiency by up to 80% in improving the accuracy of vision recognition services. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper makes DNN inference more sustainable by changing how big and accurate models are used throughout the day, depending on how much energy is being used. They developed a special algorithm for this that uses bigger models when energy usage is low and smaller ones when it’s high. They also created a new way to measure how well this approach works, called carbon-emission efficiency. This shows that their method can make vision recognition services up to 80% more accurate while using less energy. |
Keywords
* Artificial intelligence * Inference
