<s>
The	O
junction	B-Algorithm
tree	I-Algorithm
algorithm	I-Algorithm
(	O
also	O
known	O
as	O
'	O
Clique	O
Tree	O
 '	O
)	O
is	O
a	O
method	O
used	O
in	O
machine	O
learning	O
to	O
extract	O
marginalization	O
in	O
general	O
graphs	O
.	O
</s>
<s>
Performing	O
this	O
algorithm	O
for	O
a	O
graph	O
with	O
treewidth	O
k	O
will	O
thus	O
have	O
at	O
least	O
one	O
computation	O
which	O
takes	O
time	O
exponential	O
in	O
k	O
.	O
It	O
is	O
a	O
message	B-Operating_System
passing	I-Operating_System
algorithm	O
.	O
</s>
<s>
Thus	O
,	O
by	O
triangulating	B-Algorithm
a	O
graph	O
,	O
we	O
make	O
sure	O
that	O
the	O
corresponding	O
junction	O
tree	O
exists	O
.	O
</s>
<s>
This	O
can	O
be	O
efficiently	O
done	O
by	O
,	O
for	O
example	O
,	O
modifying	O
Kruskal	B-Algorithm
's	I-Algorithm
algorithm	I-Algorithm
.	O
</s>
<s>
A	O
specific	O
use	O
could	O
be	O
found	O
in	O
auto	B-Algorithm
encoders	I-Algorithm
,	O
which	O
combine	O
the	O
graph	O
and	O
a	O
passing	O
network	O
on	O
a	O
large	O
scale	O
automatically	O
.	O
</s>
<s>
It	O
is	O
an	O
approximate	B-General_Concept
inference	I-General_Concept
.	O
</s>
<s>
Cutset	B-Algorithm
conditioning	O
:	O
Used	O
with	O
smaller	O
sets	O
of	O
variables	O
.	O
</s>
<s>
Cutset	B-Algorithm
conditioning	O
allows	O
for	O
simpler	O
graphs	O
that	O
are	O
easier	O
to	O
read	O
but	O
are	O
not	O
exact	O
.	O
</s>
