文章摘要
尚青霞,周磊,冯亮.基于降噪自动编码器的多任务优化算法[J].,2019,59(4):417-426
基于降噪自动编码器的多任务优化算法
Multi-task optimization algorithm based on denoising auto-encoder
  
DOI:10.7511/dllgxb201904013
中文关键词: 多任务优化(MTO)  多任务学习(MTL)  降噪自动编码器  单任务优化  基于种群的搜索算法
英文关键词: multi-task optimization(MTO)  multi-task learning(MTL)  denoising auto-encoder  single-task optimization  population-based search algorithm
基金项目:国家自然科学基金青年基金资助项目(61603064);中央高校前沿交叉项目(106112017CDJQJ188828);重庆市前沿基础应用面上项目(cstc2017jcyjAX0319).
作者单位
尚青霞,周磊,冯亮  
摘要点击次数: 36
全文下载次数: 26
中文摘要:
      人类通常可同时进行多个任务的学习,将从一个任务中获得的知识应用到另一个任务中以加速此任务的学习.受此学习行为的启发,多任务学习(MTL)被提出并被广泛研究.与MTL动机类似,多任务优化(MTO)是在传统基于单任务优化算法基础上被提出的一种新型优化算法,该算法旨在同时在线执行多个任务,从一个任务中获取知识以帮助另一个任务,并进行任务间知识迁移,以提高多任务的优化性能.基于降噪自动编码器提出了一种新型MTO算法,推演出一种具有闭式解的降噪自动编码器,并利用此编码器显式地对多任务构建任务映射,从而使所提MTO算法能够利用不同的基于单任务优化算法的搜索偏好.采用常用的MTO基准进行综合性实验,验证了所提算法的有效性.
英文摘要:
      Inspired by the remarkable ability of human learning which is able to perform multiple tasks simultaneously and apply the knowledge gained from one task to help another, multi-task learning (MTL) has been proposed and well-studied in the literature. With similar motivation as MTL, multi-task optimization (MTO) has recently been proposed as a new algorithm for optimization. In contrast to the traditional single-task optimization algorithm, MTO conducts the optimization process on multiple problems simultaneously. It aims to improve MTO performance across multiple problems by seamlessly transferring knowledge between them online. A new MTO algorithm is proposed based on denoising auto-encoder. A denoising auto-encoder is derived for building mappings across closed-form solution, thus making the proposed MTO algorithm be able to use the search preferences induced by different single-task optimization. To evaluate the validity of the proposed algorithm, comprehensive empirical studies on complex MTO benchmark sets have been presented.
查看全文   查看/发表评论  下载PDF阅读器
关闭