EC-NAS1
收藏arXiv2024-03-22 更新2024-06-21 收录
下载链接：
https://github.com/saintslab/EC-NAS-Bench
下载链接
链接失效反馈官方服务：
资源简介：
EC-NAS1数据集是由哥本哈根大学计算机科学系创建，专注于神经架构搜索（NAS）中的能源消耗问题。该数据集包含超过160万条记录，涵盖了多种神经网络架构的能源消耗数据。创建过程中，研究团队利用了代理模型来预测能源消耗，从而减少了数据集构建的能源开销。EC-NAS1数据集的应用领域主要集中在通过多目标优化算法，探索能源效率与模型性能之间的平衡，旨在发现既节能又高效的深度学习模型。
The EC-NAS1 dataset was created by the Department of Computer Science, University of Copenhagen, focusing on the energy consumption issue in Neural Architecture Search (NAS). This dataset contains over 1.6 million records covering energy consumption data of various neural network architectures. During its construction, the research team utilized proxy models to predict energy consumption, thereby reducing the energy overhead of dataset building. The application fields of the EC-NAS1 dataset mainly focus on exploring the balance between energy efficiency and model performance via multi-objective optimization algorithms, aiming to discover deep learning models that are both energy-saving and high-performance.
提供机构：
哥本哈根大学计算机科学系
创建时间：
2022-10-12
