Topics (high frequent terms):
名次 |
微博 |
Pagerank |
李娜 |
1 |
0.5158 |
morgana牟丛 |
2 |
0.1263 |
冯喆小胖 |
3 |
0.0248 |
电商圈小七说 |
4 |
0.0130 |
易建联 |
5 |
0.0110 |
刘璇 |
6 |
0.0092 |
…… |
网络结构 \(\rightarrow\) 结构结点(structure vertex) |
文本信息 \(\rightarrow\) 属性结点(attribute vertex) |
\[ \left(\begin{array}{cccc} P_s & B \\ A & O \end{array}\right) \]
\(P_s\) |
原结构结点转移矩阵\(\times\alpha\) |
\(A\) |
结构点到属性点的转移矩阵(t(doc-topic矩阵)\(\times\beta\)) |
\(B\) |
属性点到结构点的转移矩阵(列归一化doc-topic矩阵) |
\(O\) |
零矩阵 |
\[{\rm PageRank}(p_i) = \frac{1-d}{N} + d \sum_{p_j \in M(p_i)} \frac{{\rm PageRank} (p_j)}{L(p_j)}\]
structure \(\rightarrow\) structure |
\(d_1\) |
structure \(\rightarrow\) attribute |
\(d_2\) |
attribute \(\rightarrow\)structure |
\(\alpha{d_1}+\beta{d_2}\) |
Assumption: the sum of all PageRanks is one |
\[PR= \left(\begin{array}{cccc} \frac{(1-d_1)\alpha}{N_s} \\ \frac{(1-d_1)\alpha}{N_s} \\ \dots \\ \frac{(1-d_2)\beta}{N_b} \\ \frac{(1-d_2)\beta}{N_b} \end{array}\right) + \left(\begin{array}{cccc} p_s & b \\ a & O \end{array}\right)\times{PR} \]
Top Topics: