<s>
The	O
in-crowd	B-Algorithm
algorithm	I-Algorithm
is	O
a	O
numerical	O
method	O
for	O
solving	O
basis	O
pursuit	O
denoising	O
quickly	O
;	O
faster	O
than	O
any	O
other	O
algorithm	O
for	O
large	O
,	O
sparse	O
problems	O
.	O
</s>
<s>
This	O
algorithm	O
is	O
an	O
active	B-Algorithm
set	I-Algorithm
method	I-Algorithm
,	O
which	O
minimizes	O
iteratively	O
sub-problems	O
of	O
the	O
global	O
basis	O
pursuit	O
denoising	O
:	O
</s>
<s>
The	O
active	B-Algorithm
set	I-Algorithm
strategies	O
are	O
very	O
efficient	O
in	O
this	O
context	O
as	O
only	O
few	O
coefficient	O
are	O
expected	O
to	O
be	O
non-zero	O
.	O
</s>
<s>
Other	O
active-set	B-Algorithm
methods	I-Algorithm
for	O
the	O
basis	O
pursuit	O
denoising	O
includes	O
BLITZ	O
,	O
where	O
the	O
selection	O
of	O
the	O
active	B-Algorithm
set	I-Algorithm
is	O
performed	O
using	O
the	O
duality	B-Algorithm
gap	I-Algorithm
of	O
the	O
problem	O
,	O
and	O
The	O
Feature	O
Sign	O
Search	O
,	O
where	O
the	O
features	O
are	O
included	O
based	O
on	O
the	O
estimate	O
of	O
their	O
sign	O
.	O
</s>
<s>
Since	O
every	O
time	O
the	O
in-crowd	B-Algorithm
algorithm	I-Algorithm
performs	O
a	O
global	O
search	O
it	O
adds	O
up	O
to	O
components	O
to	O
the	O
active	B-Algorithm
set	I-Algorithm
,	O
it	O
can	O
be	O
a	O
factor	O
of	O
faster	O
than	O
the	O
best	O
alternative	O
algorithms	O
when	O
this	O
search	O
is	O
computationally	O
expensive	O
.	O
</s>
<s>
A	O
theorem	O
guarantees	O
that	O
the	O
global	O
optimum	O
is	O
reached	O
in	O
spite	O
of	O
the	O
many-at-a-time	O
nature	O
of	O
the	O
in-crowd	B-Algorithm
algorithm	I-Algorithm
.	O
</s>
