Shared Structure Learning for Multiple Tasks with Multiple Views Xin Jin 1,2 , Fuzhen Zhuang 1 , Shuhui Wang 1 , Qing He 1 , and Zhongzhi Shi 1 1 Key Laboratory of Intelligent Information Processing, Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100190, China 2 University of Chinese Academy of Sciences, Beijing 100049, China {jinx,zhuangfz,heq,shizz}@ics.ict.ac.cn, wangshuhui@ict.ac.cn Abstract. Real-world problems usually exhibit dual-heterogeneity, i.e., every task in the problem has features from multiple views, and multi- ple tasks are related with each other through one or more shared views. To solve these multi-task problems with multiple views, we propose a shared structure learning framework, which can learn shared predictive structures on common views from multiple related tasks, and use the con- sistency among different views to improve the performance. An alternat- ing optimization algorithm is derived to solve the proposed framework. Moreover, the computation load can be dealt with locally in each task during the optimization, through only sharing some statistics, which sig- nificantly reduces the time complexity and space complexity. Experimen- tal studies on four real-world data sets demonstrate that our framework significantly outperforms the state-of-the-art baselines. Keywords: Multi-task Learning, Multi-view Learning, Alternating Optimization. 1 Introduction In many practical situations, people need to solve a number of related tasks, and multi-task learning (MTL) [5–7, 20] is a good choice for these problems. It learns multiple related tasks together so as to improve the performance of each task relative to learning them separately. Besides, many problems contain different “kinds” of information, that is, they include multi-view data. Multi-view learning (MVL) [3, 8, 19] can make better use of these different views and get improved results. However, many real-world problems exhibit dual-heterogeneity [14]. To be specific, a single learning task might have features in multiple views (i.e., feature heterogeneity); different learning tasks might be related with each other through one or more shared views (features) (i.e., task heterogeneity). One example is the web page classification problem. If people want to identify whether the web pages from different universities are course home pages, then classifying each university can be seen as a task. Meanwhile, every web page has different kinds of features, one kind is the content of the web page, and the H. Blockeel et al. (Eds.): ECML PKDD 2013, Part II, LNAI 8189, pp. 353–368, 2013. c  Springer-Verlag Berlin Heidelberg 2013