| - 分享主题：Task Grouping - 论文标题：Robust Task Grouping with Representative Tasks for Clustered Multi-Task Learning - 论文链接：https://dl.acm.org/doi/pdf/10.1145/3292500.3330904 - 分享人：唐共勇 |
|---|
1. Summary
【必写】，推荐使用 grammarly 检查语法问题，尽量参考论文 introduction 的写作方式。需要写出
- 这篇文章解决了什么问题？
- 作者使用了什么方法（不用太细节）来解决了这个问题？
- 你觉得你需要继续去研究哪些概念才会加深你对这篇文章的理解？
In this paper, a robust clustering multitask learning method is proposed. The tasks are grouped into several groups by learning representative tasks. The main assumption behind this method is that each task can be represented by a linear combination of some representative tasks, which can represent all tasks. The correlation coefficient between tasks can be expressed by the corresponding combination coefficient. By imposing row sparsity constraints on the correlation matrix, this method can select representative tasks and encourage information sharing among related tasks. In addition, l1,2-norm is applied to represent the loss to enhance the robustness of the method. In order to solve the resulting biconvex optimization problem, an effective optimization method based on multiplier alternating direction method and accelerated proximal gradient method is designed.
2. 你对于论文的思考
需要写出你自己对于论文的思考，例如优缺点，你的takeaways
多任务之间的分组（通过聚类的方式），指出哪几个任务之间应共享信息
任务聚类：选出一组代表任务，这组代表任务通过线性组合的方式可以表示所有任务
其优点有：
若有收获，就点个赞吧
0 人点赞
