@article{__nodate,
	title = {基于牵引控制的深度增强学习路由策略生成},
	issn = {1000-1239},
	url = {https://kns.cnki.net/kcms/detail/detail.aspx?dbcode=CAPJ&dbname=CAPJDAY&filename=JFYZ20210125001&v=8SqcXNnYzq2EiT7cTJlOWMKj49%25mmd2BvIBxSfi7RhYLeWUAi%25mmd2FBzh8o8qRxslDIo%25mmd2BUr0F},
	abstract = {当前网络规模的高速增长带来网络流量复杂度的日益提高，增加了对流量特征精确建模的难度.近年来业界提出使用深度增强学习技术实现网络路由的智能化生成，一定程度上克服了人工进行流量分析和建模的缺点.然而，目前提出的解决方案普遍存在可扩展性差等问题.对此，提出了一种基于牵引控制理论的深度增强学习路由策略生成技术Hierar-DRL，通过引入牵引控制理论并结合深度增强学习的自动策略搜索能力，提高了智能路由算法可扩展性.仿真实验结果表明，所提方案相比当前最优方案的端到端时延最多降低了28.5\%，证明了所提智能路由方案的有效性.},
	language = {中文},
	urldate = {2021-01-26},
	journal = {计算机研究与发展},
	author = {孙, 鹏浩 and 兰, 巨龙 and 申, 涓 and 胡, 宇翔},
	keywords = {artificial intelligence, deep reinforcement learning, pinning control, routing optimization, software-defined networking, 人工智能, 深度增强学习, 牵引控制, 路由优化, 软件定义网络},
	pages = {1--13},
}

{"_id":"NbbQk9yfm8EcZMpy3","bibbaseid":"----","downloads":0,"creationDate":"2016-03-05T09:20:05.717Z","title":"基于牵引控制的深度增强学习路由策略生成","author_short":["孙, 鹏.","兰, 巨.","申, 涓","胡, 宇."],"year":null,"bibtype":"article","biburl":"https://bibbase.org/zotero/ericmei","bibdata":{"bibtype":"article","type":"article","title":"基于牵引控制的深度增强学习路由策略生成","issn":"1000-1239","url":"https://kns.cnki.net/kcms/detail/detail.aspx?dbcode=CAPJ&dbname=CAPJDAY&filename=JFYZ20210125001&v=8SqcXNnYzq2EiT7cTJlOWMKj49%25mmd2BvIBxSfi7RhYLeWUAi%25mmd2FBzh8o8qRxslDIo%25mmd2BUr0F","abstract":"当前网络规模的高速增长带来网络流量复杂度的日益提高，增加了对流量特征精确建模的难度.近年来业界提出使用深度增强学习技术实现网络路由的智能化生成，一定程度上克服了人工进行流量分析和建模的缺点.然而，目前提出的解决方案普遍存在可扩展性差等问题.对此，提出了一种基于牵引控制理论的深度增强学习路由策略生成技术Hierar-DRL，通过引入牵引控制理论并结合深度增强学习的自动策略搜索能力，提高了智能路由算法可扩展性.仿真实验结果表明，所提方案相比当前最优方案的端到端时延最多降低了28.5%，证明了所提智能路由方案的有效性.","language":"中文","urldate":"2021-01-26","journal":"计算机研究与发展","author":[{"propositions":[],"lastnames":["孙"],"firstnames":["鹏浩"],"suffixes":[]},{"propositions":[],"lastnames":["兰"],"firstnames":["巨龙"],"suffixes":[]},{"propositions":[],"lastnames":["申"],"firstnames":["涓"],"suffixes":[]},{"propositions":[],"lastnames":["胡"],"firstnames":["宇翔"],"suffixes":[]}],"keywords":"artificial intelligence, deep reinforcement learning, pinning control, routing optimization, software-defined networking, 人工智能, 深度增强学习, 牵引控制, 路由优化, 软件定义网络","pages":"1–13","bibtex":"@article{__nodate,\n\ttitle = {基于牵引控制的深度增强学习路由策略生成},\n\tissn = {1000-1239},\n\turl = {https://kns.cnki.net/kcms/detail/detail.aspx?dbcode=CAPJ&dbname=CAPJDAY&filename=JFYZ20210125001&v=8SqcXNnYzq2EiT7cTJlOWMKj49%25mmd2BvIBxSfi7RhYLeWUAi%25mmd2FBzh8o8qRxslDIo%25mmd2BUr0F},\n\tabstract = {当前网络规模的高速增长带来网络流量复杂度的日益提高，增加了对流量特征精确建模的难度.近年来业界提出使用深度增强学习技术实现网络路由的智能化生成，一定程度上克服了人工进行流量分析和建模的缺点.然而，目前提出的解决方案普遍存在可扩展性差等问题.对此，提出了一种基于牵引控制理论的深度增强学习路由策略生成技术Hierar-DRL，通过引入牵引控制理论并结合深度增强学习的自动策略搜索能力，提高了智能路由算法可扩展性.仿真实验结果表明，所提方案相比当前最优方案的端到端时延最多降低了28.5\\%，证明了所提智能路由方案的有效性.},\n\tlanguage = {中文},\n\turldate = {2021-01-26},\n\tjournal = {计算机研究与发展},\n\tauthor = {孙, 鹏浩 and 兰, 巨龙 and 申, 涓 and 胡, 宇翔},\n\tkeywords = {artificial intelligence, deep reinforcement learning, pinning control, routing optimization, software-defined networking, 人工智能, 深度增强学习, 牵引控制, 路由优化, 软件定义网络},\n\tpages = {1--13},\n}\n\n\n\n","author_short":["孙, 鹏.","兰, 巨.","申, 涓","胡, 宇."],"key":"__nodate-1","id":"__nodate-1","bibbaseid":"----","role":"author","urls":{"Paper":"https://kns.cnki.net/kcms/detail/detail.aspx?dbcode=CAPJ&dbname=CAPJDAY&filename=JFYZ20210125001&v=8SqcXNnYzq2EiT7cTJlOWMKj49%25mmd2BvIBxSfi7RhYLeWUAi%25mmd2FBzh8o8qRxslDIo%25mmd2BUr0F"},"keyword":["artificial intelligence","deep reinforcement learning","pinning control","routing optimization","software-defined networking","人工智能","深度增强学习","牵引控制","路由优化","软件定义网络"],"metadata":{"authorlinks":{}},"downloads":0,"html":""},"search_terms":["孙","兰","申","胡"],"keywords":["artificial intelligence","deep reinforcement learning","pinning control","routing optimization","software-defined networking","人工智能","深度增强学习","牵引控制","路由优化","软件定义网络"],"authorIDs":[],"dataSources":["rKju5xkNx4Gxav8Qb","TJkbwzD8s2wCxBy6Y","CmHEoydhafhbkXXt5","sWFw7a3qEGqRuCokd","rW3ueZiGjcZYmSYnb"]}