Search results

Search-based structured prediction methods have shown promising successes in both computer vision and natural language processing recently. However, most existing search-based approaches lead to a complex multi-stage learning process, which is ill-suited for scene labeling probelms with a high-dimensional output space. In this paper, a stacked learning to search method is proposed to address scene labeling tasks. We design a simplified search process consisting of a sequence of ranking fucntions, which are learned based on a stacked learning strategy to prevent over-fitting. Our method is able to encode rich prior knowledge by incorporating a variety of local and global scene features. In addition, we estimate a labeling confidence map to further improve the search efficiency from two aspects: first, it constrains the search space more effectively by pruning out low-quality solutions based on confidence scores and second, we employ the confidence map as an additional ranking feature to improve its prediction performance and thus reduce the search steps. Our approach is evaluated on both semantic segmentation and geometric labeling tasks, including the Stanford Background, Sift Flow, Geometric Context, and NYUv2 RGB-D data set. The competitive results demonstrate that our stacked learning to search method provides and effective alternative paradigm for scene labeling. ; This work was supported by the National Natural Science Foundation of China under Grant 61571026. The work of F. Cheng was supported by the China Scholarship Council. The work of X. He was supported in part by the Australian Government through the Department of Communications and in part by the Australian Research Council through the ICT Center of Excellence Program

AbstractResearch QuestionDisputes over the corporate governance impacts of common ownership continue. Differentiating from existing studies, we focus on the Chinese stock market, exploiting the Top 10 Shareholding File, which includes various investors besides institutional investors, to study the impact of common ownership built through blockholders on corporate risk‐taking behavior.Research FindingsWe find that firms with higher common ownership are less likely to engage in corporate risk‐taking, with concomitant decreases in future growth rates. Mechanism analysis shows that blockholders' common ownership exerts its influence through increasing market concentration, with concomitant lessening of market competition. Interestingly, further analyses indicate that, in contrast to blockholders, ownership connectedness built by mutual fund families significantly raises corporate risk‐taking along with growth. However, individual investors' common ownership does not show the significant statistical relationship with corporate risk‐taking.Theoretical ImplicationsWe add to the debate on common ownership on corporate governance. Consistent with the anti‐competition stream of literature, the risk‐taking‐reduction role we identify for blockholder common ownership supports the theory of anti‐competition. Our results highlight the need to consider the heterogeneity of common ownership.Policy ImplicationsWhile blockholder common ownership is evidenced to have a negative effect on corporate risk‐taking, with, by extension, a negative impact on economic development, our results also suggest that efficient monitoring mitigates these effects. We also document an interesting heterogeneity in investor types. Mutual fund common ownership, in contrast to blockholder common ownership, is associated with higher risk‐taking and more robust firm growth. This suggests the positive role of institutions in corporate governance and the necessity of considering the heterogeneity of common ownership.