基于端到端分布式框架的符号网络预测方法

社交网络中的链接关系根据其潜在含义可分为正关系和负关系。若对网络中的链接关系进行正负标注，则可形成一个符号网络。符号网络在社会学、信息学、生物学等多个领域存在广泛应用。针对符号网络中链接关系的正负预测已经成为当前研究的热点之一。在大数据背景下，随着符号网络规模的日益增大，符号预测算法的可伸缩性问题日益突出。一些研究者提出了分布式环境下的符号预测方法，使得算法的可伸缩性问题部分得到缓解。但由于大多数算法采用了服务器-客户端方式的分布式框架，导致问题并没有得到根本上的解决。本文提出了一种新的端到端分布式框架（Client to Client Distributed FrameWork，简称C2CDF），相比传统服务器-客户端架构的集中通信模式，C2CDF各个节点间地位平等，不存在集中通信，集群的带宽瓶颈和压力大大减轻。通过在社交网络正负符号预测、广告点击率预测及森林类型预测等三个不同真实数据集上的实验证明，C2CDF能够在拥有更高准确性的同时，获得2.3-3.3倍的加速比，而且拥有良好的泛化性，不仅能应用在社交网络正负符号预测方面，也能作用于广告点击预测等其他领域。

Signed Network Prediction Method Based on the Client-to-Client Distributed Framework

The edges of a network can be divided into positive and negative relationships according to their potential meanings. If the edges of a network are signed with plusor minus signsrepectively, a signed network can be formed. Signed networks are widely used in many fields such as sociology, informatics, biology and so on. So, the sign predictionproblem in signed networks has become one of research hotspots.In the large amount of dateset, the scalability of sign prediction algorithm is still a great challenge. There are many related works in the distributed design of signed network prediction methods, however,the computation effciency is still limited by the foundational master-slave framework. In this paper, we propose C2CDF (Client to Client Distributed Framework). Compared with traditional master-slave framework,C2CDF is a completely new client-to-client frameworkwhich can release the bandwidth pressure by abandoning the server node and allowing the communications between the client nodes. The Experiments,on sign prediction insigned social networks, prediction inclick-through rate and prediction in forest type, show that C2CDFis a general approach which not only can be applied in sign prediction in signed network, but also be used in the other prediction area. In these three datasets, C2CDF can achieve better performance than FM inferred by the traditional SGD algorithm. We also demonstrate that C2CDFachieves a 2.3-3.3x speed-up over the method implementedunder the master-slave framework, while obtains a better accuracy performance than the comparison method.