Graphics Reference
In-Depth Information
IB2 is similar to CNN, except that IB2 not seed S with one instance of each class
and does not repeat the process after the first pass through the training set.
IB3
-The IB3 algorithm proceeds as follows:
For each instance
X
i
in
TR
Let
a
be the nearest acceptable instance in
S
to
X
i
.
(if there are no acceptable instances in
S
,
let
a
be a random instance in
S
)
If
class
X
i
)
then add
X
i
to
S
.
For each instance
s
in
S
If
s
is at least as close to
X
i
as
a
is
Update the classification record of
s
and remove
s
from
S
its classification
record is significantly poor.
Remove all non-acceptable instance from
S
.
An instance is
acceptable
if the lower bound of its accuracy is statistically signifi-
cantly higher (at a 90 % confidence level) than the upper bound on the frequency of
its class. Similarly, an instance is dropped from
S
if the upper bound on its accuracy
is statistically significantly lower (at a 70 % confidence level) than the lower bound
on the frequency of its class. Other instances are kept in
S
during training, and then
dropped at the end if they do not prove to be acceptable.
The expression for the upper and lower bounds of the confidence level interval is:
(
a
)
=
class
(
z
p
(
1
−
p
)
g
z
2
4
g
2
z
2
+
/
±
+
p
2
g
(8.3)
z
2
1
+
/
g
where for the
accuracy
of an instance in
S
,
g
is the number of classification attempts
since introduction of the instance to
S
(i.e., the number of times it was at least as
close to
X
i
as
a
was),
p
is the accuracy of such attempts (i.e., the number of times
the instance's class matched
X
i
's class, divided by
g
), and
z
is the confidence (0.9 for
acceptance, 0.7 for dropping). For the frequency of a class,
p
is the frequency (i.e.
proportion of instances so far that are of this class),
g
is the number of previously
processed instances, and
z
is the confidence.
8.4.3.2 Decremental
8.4.3.3 Filter
•
Variable Similarity Metric (VSM)
[
111
]—In order to reduce storage and remove
noisy instances, an instance
X
i
is removed if all
k
of its neighbors are of the same