<s>
struc2vec	B-General_Concept
is	O
a	O
framework	O
to	O
generate	O
node	O
vector	O
representations	O
on	O
a	O
graph	O
that	O
preserve	O
the	O
structural	O
identity	O
.	O
</s>
<s>
In	O
contrast	O
to	O
node2vec	B-General_Concept
,	O
that	O
optimizes	O
node	O
embeddings	O
so	O
that	O
nearby	O
nodes	O
in	O
the	O
graph	O
have	O
similar	O
embedding	O
,	O
struc2vec	B-General_Concept
captures	O
the	O
roles	O
of	O
nodes	O
in	O
a	O
graph	O
,	O
even	O
if	O
structurally	O
similar	O
nodes	O
are	O
far	O
apart	O
in	O
the	O
graph	O
.	O
</s>
<s>
struc2vec	B-General_Concept
identifies	O
nodes	O
that	O
play	O
a	O
similar	O
role	O
based	O
solely	O
on	O
the	O
structure	O
of	O
the	O
graph	O
,	O
for	O
example	O
computing	O
the	O
structural	O
identity	O
of	O
individuals	O
in	O
social	O
networks	O
.	O
</s>
<s>
In	O
particular	O
,	O
struc2vec	B-General_Concept
employs	O
a	O
degree-based	O
method	O
to	O
measure	O
the	O
pairwise	O
structural	O
role	O
similarity	O
,	O
which	O
is	O
then	O
adopted	O
to	O
build	O
the	O
multi-layer	O
graph	O
.	O
</s>
<s>
struc2vec	B-General_Concept
follows	O
the	O
intuition	O
that	O
random	O
walks	O
through	O
a	O
graph	O
can	O
be	O
treated	O
as	O
sentences	O
in	O
a	O
corpus	O
.	O
</s>
<s>
In	O
its	O
final	O
phase	O
,	O
the	O
algorithm	O
employs	O
Gensim	B-Application
's	O
word2vec	B-Algorithm
algorithm	O
to	O
learn	O
embeddings	O
based	O
on	O
biased	O
random	O
walks	O
.	O
</s>
<s>
Sequences	O
of	O
nodes	O
are	O
fed	O
into	O
a	O
skip-gram	B-Algorithm
or	O
continuous	B-Algorithm
bag	I-Algorithm
of	I-Algorithm
words	I-Algorithm
model	O
and	O
traditional	O
machine-learning	O
techniques	O
for	O
classification	O
can	O
be	O
used	O
.	O
</s>
