<s>
Within	O
artificial	B-Application
intelligence	I-Application
and	O
operations	O
research	O
for	O
constraint	B-Application
satisfaction	I-Application
a	O
hybrid	B-Application
algorithm	I-Application
solves	O
a	O
constraint	B-Application
satisfaction	I-Application
problem	I-Application
by	O
the	O
combination	O
of	O
two	O
different	O
methods	O
,	O
for	O
example	O
variable	O
conditioning	O
(	O
backtracking	B-Algorithm
,	O
backjumping	B-Application
,	O
etc	O
.	O
)	O
</s>
<s>
Hybrid	B-Application
algorithms	I-Application
exploit	O
the	O
good	O
properties	O
of	O
different	O
methods	O
by	O
applying	O
them	O
to	O
problems	O
they	O
can	O
efficiently	O
solve	O
.	O
</s>
<s>
This	O
hybrid	B-Application
algorithm	I-Application
is	O
based	O
on	O
running	O
search	O
over	O
a	O
set	O
of	O
variables	O
and	O
inference	O
over	O
the	O
other	O
ones	O
.	O
</s>
<s>
In	O
particular	O
,	O
backtracking	B-Algorithm
or	O
some	O
other	O
form	O
of	O
search	O
is	O
run	O
over	O
a	O
number	O
of	O
variables	O
;	O
whenever	O
a	O
consistent	O
partial	O
assignment	O
over	O
these	O
variables	O
is	O
found	O
,	O
inference	O
is	O
run	O
over	O
the	O
remaining	O
variables	O
to	O
check	O
whether	O
this	O
partial	O
assignment	O
can	O
be	O
extended	O
to	O
form	O
a	O
solution	O
.	O
</s>
<s>
Another	O
hybrid	O
search/inference	O
algorithm	O
works	O
on	O
the	O
tree	B-Application
decomposition	I-Application
.	O
</s>
<s>
In	O
general	O
,	O
a	O
constraint	B-Application
satisfaction	I-Application
problem	I-Application
can	O
be	O
solved	O
by	O
first	O
creating	O
a	O
tree	B-Application
decomposition	I-Application
and	O
then	O
using	O
a	O
specialized	O
algorithm	O
.	O
</s>
<s>
A	O
hybrid	O
approach	O
can	O
be	O
taken	O
by	O
using	O
variable	O
elimination	O
for	O
creating	O
the	O
new	O
constraints	O
that	O
are	O
propagated	O
within	O
nodes	O
,	O
and	O
a	O
search	O
algorithm	O
(	O
such	O
as	O
backtracking	B-Algorithm
,	O
backjumping	B-Application
,	O
local	B-Application
search	I-Application
)	O
on	O
each	O
individual	O
node	O
.	O
</s>
