图的常见应用
link prediction
已知现在的图结构，predict future link/missing link
启发式
common neighbors：相同的邻居越多，有link的概率越大
CN
(
x
,
y
)
=
∣
Γ
(
x
)
∩
Γ
(
y
)
∣
\text{CN}(x,y) = |\Gamma(x) \cap \Gamma(y)|
CN(x,y)=∣Γ(x)∩Γ(y)∣
jaccard coefficient：
JC
(
x
,
y
)
=
∣
Γ
(
x
)
∩
Γ
(
y
)
∣
∣
Γ
(
x
)
∪
Γ
(
y
)
∣
\text{JC}(x, y) = \frac{|\Gamma(x) \cap \Gamma(y)|}{|\Gamma(x) \cup \Gamma(y)|}
JC(x,y)=∣Γ(x)∪Γ(y)∣∣Γ(x)∩Γ(y)∣
Hub promoted：用两个节点的最小邻居节点数做个归一化
HP
(
x
,
y
)
=
∣
Γ
(
x
)
∩
Γ
(
y
)
∣
min
(
∣
Γ
(
x
)
∣
,
∣
Γ
(
y
)
∣
)
\text{HP}(x, y) = \frac{|\Gamma(x) \cap \Gamma(y)|}{\min(|\Gamma(x)|, |\Gamma(y)|)}
HP(x,y)=min(∣Γ(x)∣,∣Γ(y)∣)∣Γ(x)∩Γ(y)∣
admic adar：对频繁与其他节点common的common node降权
AA
(
x
,
y
)
=
∑
z
∈
Γ
(
x
)
∩
Γ
(
y
)
1
log
∣
Γ
(
z
)
∣
\text{AA}(x,y) = \sum_{z \in \Gamma(x) \cap \Gamma(y)} \frac{1}{\log |\Gamma(z)|}
AA(x,y)=z∈Γ(x)∩Γ(y)∑log∣Γ(z)∣1
ressource allocation：基本同admic adar
RA
(
x
,
y
)
=
∑
z
∈
Γ
(
x
)
∩
Γ
(
y
)
1
∣
Γ
(
z
)
∣
\text{RA}(x,y) = \sum_{z \in \Gamma(x) \cap \Gamma(y)} \frac{1}{|\Gamma(z)|}
RA(x,y)=z∈Γ(x)∩Γ(y)∑∣Γ(z)∣1
preferential attachment：只考虑两个节点的邻居节点个数
PA
(
x
,
y
)
=
∣
Γ
(
x
)
∣
⋅
∣
Γ
(
y
)
∣
\text{PA}(x,y) = |\Gamma(x)| \cdot |\Gamma(y)|
PA(x,y)=∣Γ(x)∣⋅∣Γ(y)∣
ml
一般可以用上面的启发式指标作为feature
node classification
默认图的边没有缺失，给节点分类
扫一扫
4349
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
