报告题目：

Doubly Dividing the Massive Data for Prediction Using Model Aggregation

报 告 人：

吴远山 副教授（37000cm威尼斯）

报告时间：

2017年11月30日 12:55--13:30

报告地点：

数学院东北楼一楼报告厅（110）

报告摘要：

We propose a doubly dividing model aggregation method to enhance the prediction accuracy based on the massive data, which are typically storied in a distributed manner. We further devise a communication-efficient algorithm to resolve the optimal weights for each working model. Within several rounds of communications, we show that our method can achieve prediction error bound of the oracle method. Compared with the existing methods, the proposed method delivers favorable performance in numerical studies