Journal of Frontiers of Computer Science and Technology ›› 2017, Vol. 11 ›› Issue (6): 988-997.DOI: 10.3778/j.issn.1673-9418.1603067

Previous Articles     Next Articles

Fast Algorithm for Regularized Multi-Task Learning

SHI Yingzhong1,2+, WANG Juqin1,2, XU Min1, WANG Shitong1   

  1. 1. School of Digital Media, Jiangnan University, Wuxi, Jiangsu 214122, China
    2. College of Internet of Things, Wuxi Institute of Technology, Wuxi, Jiangsu 214121, China
  • Online:2017-06-01 Published:2017-06-07


史荧中1,2+,汪菊琴1,2,许  敏1,王士同1   

  1. 1. 江南大学 数字媒体学院,江苏 无锡 214122
    2. 无锡职业技术学院 物联网学院,江苏 无锡 214121

Abstract: Regularized multi-task learning (rMTL) method and its extensions have achieved remarkable achievement in theoretical research and applications. However, previous research focuses on the relationship of tasks instead of the complexity of algorithms, the high computational cost of these methods are impractical for large scale datasets. In order to overcome this shortcoming, this paper proposes a fast regularized multi-task learning (FrMTL) based on core vector machine (CVM) theory. FrMTL is competitive with rMTL in classification accuracy while FrMTL-CVM can make a decision rapidly for large scale datasets because of the merit of asymptotic linear time complexity. The effectiveness of the proposed classifier is also experimentally confirmed.

Key words: multi-task learning, large scale dataset, core vector machine, fast classification

摘要: 正则化多任务学习(regularized multi-task learning,rMTL)方法及其扩展方法在理论研究及实际应用方面已经取得了较好的成果。然而以往方法仅关注于多个任务之间的关联,而未充分考虑算法的复杂度,较高的计算代价限制了其在大数据集上的实用性。针对此不足,结合核心向量机(core vector machine,CVM)理论,提出了适用于多任务大数据集的快速正则化多任务学习(fast regularized multi-task learning,FrMTL)方法。FrMTL方法有着与rMTL方法相当的分类性能,而基于CVM理论的FrMTL-CVM算法的渐近线性时间复杂度又能使其在面对大数据集时仍然能够获得较快的决策速度。该方法的有效性在实验中得到了验证。

关键词: 多任务学习, 大数据集, 核心向量机, 快速分类