arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2301.09231cs.LGstat.APstat.ML
收藏

GP-NAS-ensemble:一种用于NAS性能预测的模型

GP-NAS-ensemble: a model for NAS Performance Prediction

Meituan(美团);Tencent(腾讯);Bigo Technology(Bigo科技);Alibaba Group(阿里巴巴集团);Chinese Academy of Sciences(中国科学院)

Kunlong Chen, Liu Yang, Yitian Chen, Kunjin Chen, Yidan Xu, Lujun Li

摘要:

在神经网络结构搜索(Neural Architecture Search,NAS)的应用中,由于评估一个结构的性能需要花费大量的时间,因此在不进行训练的情况下对给定模型结构的性能进行评估具有重要意义。在本文中,一个新的NAS框架称为GP-NAS-ensemble提出了预测性能的神经网络架构与一个小的训练数据集。本文对GP-NAS模型进行了改进,使其具有集成学习方法的优点。我们的方法在CVPR 2022第二届轻量级NAS挑战赛性能预测赛道中排名第二。

英文摘要:

It is of great significance to estimate the performance of a given model architecture without training in the application of Neural Architecture Search (NAS) as it may take a lot of time to evaluate the performance of an architecture. In this paper, a novel NAS framework called GP-NAS-ensemble is proposed to predict the performance of a neural network architecture with a small training dataset. We make several improvements on the GP-NAS model to make it share the advantage of ensemble learning methods. Our method ranks second in the CVPR2022 second lightweight NAS challenge performance prediction track.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑