<s>
Stochastic	B-Algorithm
optimization	I-Algorithm
(	O
SO	O
)	O
methods	O
are	O
optimization	O
methods	O
that	O
generate	O
and	O
use	O
random	O
variables	O
.	O
</s>
<s>
Stochastic	B-Algorithm
optimization	I-Algorithm
methods	O
also	O
include	O
methods	O
with	O
random	O
iterates	O
.	O
</s>
<s>
Some	O
stochastic	B-Algorithm
optimization	I-Algorithm
methods	O
use	O
random	O
iterates	O
to	O
solve	O
stochastic	O
problems	O
,	O
combining	O
both	O
meanings	O
of	O
stochastic	B-Algorithm
optimization	I-Algorithm
.	O
</s>
<s>
Stochastic	B-Algorithm
optimization	I-Algorithm
methods	O
generalize	O
deterministic	O
methods	O
for	O
deterministic	O
problems	O
.	O
</s>
<s>
Partly	O
random	O
input	O
data	O
arise	O
in	O
such	O
areas	O
as	O
real-time	O
estimation	O
and	O
control	O
,	O
simulation-based	O
optimization	O
where	O
Monte	B-Algorithm
Carlo	I-Algorithm
simulations	I-Algorithm
are	O
run	O
as	O
estimates	O
of	O
an	O
actual	O
system	O
,	O
and	O
problems	O
where	O
there	O
is	O
experimental	O
(	O
random	O
)	O
error	O
in	O
the	O
measurements	O
of	O
the	O
criterion	O
.	O
</s>
<s>
On	O
the	O
other	O
hand	O
,	O
even	O
when	O
the	O
data	B-General_Concept
set	I-General_Concept
consists	O
of	O
precise	O
measurements	O
,	O
some	O
methods	O
introduce	O
randomness	O
into	O
the	O
search-process	O
to	O
accelerate	O
progress	O
.	O
</s>
<s>
Such	O
randomness	O
can	O
also	O
make	O
the	O
method	B-Algorithm
less	O
sensitive	O
to	O
modeling	O
errors	O
.	O
</s>
<s>
Further	O
,	O
the	O
injected	O
randomness	O
may	O
enable	O
the	O
method	B-Algorithm
to	O
escape	O
a	O
local	O
optimum	O
and	O
eventually	O
to	O
approach	O
a	O
global	O
optimum	O
.	O
</s>
<s>
Indeed	O
,	O
this	O
randomization	B-General_Concept
principle	O
is	O
known	O
to	O
be	O
a	O
simple	O
and	O
effective	O
way	O
to	O
obtain	O
algorithms	O
with	O
almost	O
certain	O
good	O
performance	O
uniformly	O
across	O
many	O
data	B-General_Concept
sets	I-General_Concept
,	O
for	O
many	O
sorts	O
of	O
problems	O
.	O
</s>
<s>
Stochastic	B-Algorithm
optimization	I-Algorithm
methods	O
of	O
this	O
kind	O
include	O
:	O
</s>
<s>
parallel	B-Algorithm
tempering	I-Algorithm
a.k.a.	O
</s>
<s>
In	O
contrast	O
,	O
some	O
authors	O
have	O
argued	O
that	O
randomization	B-General_Concept
can	O
only	O
improve	O
a	O
deterministic	O
algorithm	O
if	O
the	O
deterministic	O
algorithm	O
was	O
poorly	O
designed	O
in	O
the	O
first	O
place	O
.	O
</s>
<s>
The	O
way	O
in	O
which	O
results	O
of	O
stochastic	B-Algorithm
optimization	I-Algorithm
algorithms	O
are	O
usually	O
presented	O
(	O
e.g.	O
,	O
presenting	O
only	O
the	O
average	O
,	O
or	O
even	O
the	O
best	O
,	O
out	O
of	O
N	O
runs	O
without	O
any	O
mention	O
of	O
the	O
spread	O
)	O
,	O
may	O
also	O
result	O
in	O
a	O
positive	O
bias	O
towards	O
randomness	O
.	O
</s>
