Balancing the Pain and Gain of Hobnobbing-师姐论文

本文链接：https://blog.csdn.net/Ipanema_Lemon/article/details/84063621

Balancing the Pain and Gain of Hobnobbing

Utility-Based Network Building over Attributed Social Networks

摘要

Existing methods for this problem rely on link prediction that evaluates the likelihood of successful tie creation between two individuals. Such methods do not consider the social capital gained by agents, nor do they concern with the required cost of this process. In light of this limitation, we propose a utility-based network building problem, with an aim to strike a balance between the gained social capital – in the form of closeness centrality – and the cost of establishing ties.
传统链接推荐基于链接预测，未考虑链接建立带来的收益以及链接建立所需的代价。基于这种考虑，提出了基于收益的网络构建问题，目标是在建立链接和获取社会资本（以closeness centrality衡量）之间找到平衡。

关键词

Network formation; link recommendation; core/periphery structure;social capital; link prediction; centrality

1 INTRODUCTION

社会位置social position的意义
将收益与代价的概念引入，并希望提出策略能够使自身位于网络的有利位置
social capital的介绍，并提出假设每个个体有一定的用于建立链接的预算
it is reasonable to assume that each individual has a finite budget which may be distributed among their social connections. The costs reflect the levels of difficulty or uncertainty involved in tie creation, and would certainly vary among different ties.
中心度介绍
主要贡献（1）提出了utility-based network building problem（U-NB）。By creating a tie, an individual gains in terms of centrality, but at the same time pays a cost that depends on link prediction: Higher probability of tie emergence means lower costs/uncertainty to establish this tie.
（2）提出个人的网络构建策略，并可以同于带属性网络。在随机网络模型和真实世界网络中都进行了实验评估
（3）提出了一种新的网络形成模型。当所有Agent在网络中都采取文章提出的模型时，网络最终演化出现了核心/边缘结构，而且满足了很多真实世界的属性。

2 RELATED WORK

3 UTILITY-BASED NETWORK BUILDING PROBLEM (U-NB)

3.1 Problem Setup

Utility-中心度
将节点中心度的提升大小作为收益的评价。

Cost-链接预测评分的倒数（链接预测越可能，建立链接的代价就越小）
本文用personalized PageRank index作为预测的评分。
Personalized PageRank is more suitable in our context as it gives a non-zero score for any reachable pairs of nodes, extending beyond pairs with nonempty overlapped neighborhoods.
Personalized PageRank 给目标节点的所有可达的节点对都能给出一个非零的评分，很适合做代价值。

A limitation of this method is that a non-zero score is only given two pairs of nodes that are in the same connected components. If u,s are not connected by a path, PRu is evaluated to 0, making the link cost between them infinite. To alleviate this problem, we introduce another parameter χ > 0 and amend the cost function to the following smoothed cost function when the graph is not connected
用Personalized PageRank问题在于，对于不在一个连通子图里的节点对，pr值是0，无法取倒数，所以添加一个参数 $\Lambda$ 来修补，即在求出来的pr值中加上 $\Lambda$ 再求倒数。

3.2 Attributed Network

采用SAN（social attribute network）网络中有两种节点，一种是属性值节点，一种是用户节点。

4 NETWORK BUILDING STRATEGIES

4.1 三种策略

Utility Strategy(SCLS)
只注重收益，在budget内贪心地连收益最大的人，会建立更少的链接。
Cost Strategy(SPP)
只关注代价最小，不停的连接代价最小最容易建立连接的人。
Unified Strategy(SUN)
考虑收益与代价的平衡，贪心地最大化收益/代价比。
strikes a balance between utility and cost, iteratively building edges for v that maximize utility/cost ratio.