算法¶
1. Clique Percolation¶
中文名?派系过滤
假设¶
群落由 完全连通子图的重叠集合组成——a community consists of overlapping sets of fully connected subgraphs
思路¶
detects communities by searching for adjacent cliques
扩展内容¶
- 派系(Cliques)。在一个无向网络图中,“派系”指的是至少包含3个点的最大完备子图。这个概念包含3层含义:①一个派系至少包含三个点。②派系是完备的,根据完备图的定义,派系中任何两点之间都存在直接联系。③派系是“最大”的,即向这个子图中增加任何一点,将改变其“完备”的性质。
- n-派系(n-Cliques)。对于一个总图来说,如果其中的一个子图满足如下条件,就称之为n-派系:在该子图中,任何两点之间在总图中的距离(即捷径的长度)最大不超过n。从形式化角度说,令d(i,j)代表两点和n在总图中的距离,那么一个n-派系的形式化定义就是一个满足如下条件的拥有点集的子图,即:d(i,J)\le n,对于所有的,n_i,n_j\in N,来说,在总图中不存在与子图中的任何点的距离不超过n的点。
- n-宗派(n—Clan)。所谓n-宗派(n—Clan)是指满足以下条件的n-派系,即其中任何两点之间的捷径的距离都不超过n。可见,所有的n-宗派都是n-派系。
- k-丛(k-Plex)。一个k-丛就是满足下列条件的一个凝聚子群,即在这样一个子群中,每个点都至少与除了k个点之外的其他点直接相连。也就是说,当这个凝聚子群的规模为n时,其中每个点至少都与该凝聚子群中n-k个点有直接联系,即每个点的度数都至少为n—k。
某个的步骤之一¶
- begins by identifying all cliques of size k in a network, a new graph is constructed such that each vertex represents one of these k-cliques
结论¶
more like pattern matching rather than finding communities since they aim to find specific, localized structure in a network.
2. Line Graph and Link Partitioning¶
中文名?连接划分
思路¶
- partitioning links
- A node in the original graph is called overlapping if links connected to it are put in more than one cluster.
3. Local Expansion and Optimization¶
中文名?局部增广和优化
思路¶
- based on growing a natural community or a partial community
- rely on a local benefit function that characterizes the quality of a densely connected group of nodes
4. Fuzzy Detection¶
中文名?模糊检测
思路¶
quantify the strength of association between all pairs of nodes and communities
例子¶
Non-negative Matrix Factorization
5. Agent-Based and Dynamical Algorithms¶
中文名?基于啥的动态算法
思路¶
label propagation algorithm by allowing a node to have multiple labels
6. others¶
中文名——无法分类 :)
评估方法¶
1. Normalized Mutual Information¶
中文名?标准互信息