摘要
神经架构搜索的目标是通过深度学习让机器学会自动优化参数,构建最优的神经架构。以神经架构搜索发展历程为时间线,从搜索空间、搜索策略和性能评估等3个方面简述神经架构搜索的算法发展和实验条件变化。分析总结典型架构搜索算法的特点和实验条件,探讨神经架构搜索技术未来发展趋势。
The goal of neural architecture search is to make machine learn automatically optimize parameters by deep learning,and build the best neural architecture.Taking the development history of neural architecture search as the timeline,the algorithm development and the experimental condition changes are briefly discussed from search space,search strategy and performance evaluation.The algorithm characteristics and experimental conditions of typical architectures are analyzed and summarized.The development trend of neural architecture search technology is discussed.
作者
潘晓英
曹园
贾蓉
戚玉涛
PAN Xiaoying;CAO Yuan;JIA Rong;QI Yutao(School of Computer Science and Technology,Xi’an University of Posts and Telecommunications,Xi’an 710121,China;Shaanxi Key Laboratory of Networks Data Analysis and Intelligent Processing,Xi’an University of Posts and Telecommunications,Xi’an 710121,China;School of Computer Science and Technology,Xidian University,Xi’an 710071,China)
出处
《西安邮电大学学报》
2022年第4期43-63,共21页
Journal of Xi’an University of Posts and Telecommunications
基金
国家自然科学基金项目(62001380)。
关键词
自动化深度学习
机器学习
神经架构搜索
进化算法
贝叶斯优化
automated deep learning
machine learning
neural architecture search
evolutionary algorithm
Bayesian optimization
