
158 Knowledge Discovery from Data Streams
Algorithm 25: The Online Bagging Algorithm.
input: L
o
: online base Learning Algorithm
h: Set of learned models {h
1
, . . . , h
m
}
(x,y): Latest training example to arrive
begin
foreach base model h
m
∈ h do
Set k according to Poisson(1)
for i = 1 to k do
h
m
← L
o
(h
m
, (x, y))
end
method able to avoid this requirement. Each original training example may be
replicated zero, one, or more times in each bootstrap training set because the
sampling is done with replacement. Each base model is trained with k copies
of each of the original training examples where:
P (k) =
exp(−1)
k!
(10.1)
As each training example is available, and