NAAP-440
收藏arXiv2022-09-28 更新2024-06-21 收录
下载链接：
https://www.kaggle.com/datasets/talcs1/naap440
下载链接
链接失效反馈官方服务：
资源简介：
NAAP-440数据集由智能射手机构创建，包含440个神经架构，这些架构均在CIFAR10数据集上使用固定配方进行训练。数据集的创建过程涉及定义神经架构的群体，并使用相同的固定训练配方对每个架构进行训练和评估。NAAP-440数据集旨在通过使用回归算法预测架构的准确性，从而加速神经架构搜索（NAS）的研究，提高其效率。
The NAAP-440 dataset was created by the Intelligent Shooter Institution. It encompasses 440 neural architectures, all of which were trained using a fixed training protocol on the CIFAR-10 dataset. The dataset creation process involves defining a population of neural architectures, followed by training and evaluating each architecture with the same fixed training protocol. The NAAP-440 dataset aims to accelerate research on Neural Architecture Search (NAS) and improve its efficiency by employing regression algorithms to predict the accuracy of these architectures.
提供机构：
智能射手
创建时间：
2022-09-14
