欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
数据挖掘是从大量数据中提取隐含知识的过程.随着数据挖掘的广泛应用,图作为一种一般数据结构在复杂结构和它们之间相互作用建模中变得越来越重要,这使得图挖掘成为数据挖掘的一个新的热点研究方向之一.由于图分类具有许多真实的应用背景,因而图分类已成为图挖掘中重要的研究领域.目前对图分类的研究都基于一个假设：训练集和测试集都是来源于同一个分布.然而,在很多真实的应用上,训练集和测试集不一定是来自同一个分布的.在本文中,我们将学习如何运用迁移学习的方法来对图数据进行分类,并提出一个基于集成学习的算法TrGBoost,该算法能在少量有标签的图数据和大量相关的图数据集里,有效地建立一个图分类器.真实数据上的实验验证了本文算法的有效性.
英文摘要：
Nowadays,the graph classification has become an important and active research topic in the last decade,which has a wild variety of real world applications.Current research on graph classification assumes that train and test data are drawn from the same feature space and same distribution.However,in many applications,the train graph data and test data are not dawn from the same distribution.In this paper,we study the transfer learning for graph classification and propose a novel solution,TraGboost,to efficiently build a graph classification model with a few labeled graph data and amounts of relation graph data.TrGboost,which based of boosting,allows users to utilize amounts of relation labeled graph data to construct a classification model for the target graph data.Empirical studies on real-world tasks demonstrate that our method can effectively boost graph classification with transfer learning.
同期刊论文项目
同项目期刊论文
期刊信息
位置：