摘要：Multi-task or multi-view learning has found wide applications respectively, but the learning problems of multi-task and multi-view have received not too much attention, in this paper, we propose a novel framework called Multi-Task Multi-View learning based on Common and Special Features (MTMVCSF), It is generally known that, multiple views can be regarded as the various representations of the raw data, and common or special information of raw data exists in these views. As a consequence, we find a way to discover the joint latent representation of multiple views in each task. By this way, the original multiple views in a task can be replaced by a latent representation, which is composed of multiple views' common and special information. Therefore, the original multi-task and multi-view learning problem degenerates to a multi-task learning problem. And by using the relationships between different tasks can improve the performance of the learning algorithm. Experimental results on four real data sets and two synthetic data sets verify the effectiveness of MTMVCSF.

关键字：Multi-task Multi-view Clustering Latent representation

ISSN号：2161-2927

卷、期、页：页: 9410-9415

发表日期：2018-01-01

期刊分区（SCI为中科院分区）：无

收录情况：EI（工程索引）

发表期刊名称：Chinese Control Conference

通讯作者：卢润坤,连思铭

第一作者：左信,刘建伟

论文类型：会议论文

论文概要：卢润坤,左信,刘建伟,连思铭，Multi-View Learning Based on Common and Special Features in Multi-Task Scenarios，Chinese Control Conference，2018，页: 9410-9415

论文题目：Multi-View Learning Based on Common and Special Features in Multi-Task Scenarios