<s>
In	O
machine	O
learning	O
,	O
weighted	B-Algorithm
majority	I-Algorithm
algorithm	I-Algorithm
(	O
WMA	O
)	O
is	O
a	O
meta	B-General_Concept
learning	I-General_Concept
algorithm	O
used	O
to	O
construct	O
a	O
compound	O
algorithm	O
from	O
a	O
pool	B-General_Concept
of	O
prediction	O
algorithms	O
,	O
which	O
could	O
be	O
any	O
type	O
of	O
learning	O
algorithms	O
,	O
classifiers	O
,	O
or	O
even	O
real	O
human	O
experts	O
.	O
</s>
<s>
The	O
algorithm	O
assumes	O
that	O
we	O
have	O
no	O
prior	O
knowledge	O
about	O
the	O
accuracy	O
of	O
the	O
algorithms	O
in	O
the	O
pool	B-General_Concept
,	O
but	O
there	O
are	O
sufficient	O
reasons	O
to	O
believe	O
that	O
one	O
or	O
more	O
will	O
perform	O
well	O
.	O
</s>
<s>
To	O
construct	O
the	O
compound	O
algorithm	O
,	O
a	O
positive	O
weight	O
is	O
given	O
to	O
each	O
of	O
the	O
algorithms	O
in	O
the	O
pool	B-General_Concept
.	O
</s>
<s>
The	O
compound	O
algorithm	O
then	O
collects	O
weighted	O
votes	O
from	O
all	O
the	O
algorithms	O
in	O
the	O
pool	B-General_Concept
,	O
and	O
gives	O
the	O
prediction	O
that	O
has	O
a	O
higher	O
vote	O
.	O
</s>
<s>
If	O
the	O
compound	O
algorithm	O
makes	O
a	O
mistake	O
,	O
the	O
algorithms	O
in	O
the	O
pool	B-General_Concept
that	O
contributed	O
to	O
the	O
wrong	O
predicting	O
will	O
be	O
discounted	O
by	O
a	O
certain	O
ratio	O
β	O
where	O
0	O
<β<	O
1	O
.	O
</s>
<s>
There	O
are	O
many	O
variations	O
of	O
the	O
weighted	B-Algorithm
majority	I-Algorithm
algorithm	I-Algorithm
to	O
handle	O
different	O
situations	O
,	O
like	O
shifting	O
targets	O
,	O
infinite	O
pools	O
,	O
or	O
randomized	O
predictions	O
.	O
</s>
<s>
The	O
core	O
mechanism	O
remains	O
similar	O
,	O
with	O
the	O
final	O
performances	O
of	O
the	O
compound	O
algorithm	O
bounded	O
by	O
a	O
function	O
of	O
the	O
performance	O
of	O
the	O
specialist	O
(	O
best	O
performing	O
algorithm	O
)	O
in	O
the	O
pool	B-General_Concept
.	O
</s>
