Information Technology Reference
In-Depth Information
c X
n
1
2 jj
2
1 ð e i þ e i Þ
R
¼
w
jj
þ
ð 6 Þ
i
¼
subject to the following requirements:
y i
f
ð
x i Þ þ e i
y i þ e i
f
ð
x i Þ
e i
0
; e i
0
;
i
¼
1
;
2
; ...;
N
4.3 Experimental Setting for Generic Machine Learning
Strategy of Testing Reliability
Algorithm 4.1 The Basic ML Algorithm line up
Input : Parameterized policy of test domain with initial test parameters ʸ = ʸ o and we
evaluate the derivative of the deviation of scope of test and actual parameter as ∇ʸlogˀ
Set parameters for different time steps, error and deviation
for t =0 , 1 , 2 , 3 , ... do
input the sample with domain scope ˀ
and set already visited bug points to a new
time step a t .
observe the deviation of sample with error
Update basis function as transpose of new test domain after deviation as [ ˆ ] T and
thus ∇ʸlogˀ becomes [ ∇ʸlogˀ ] T
Compute natural gradient and update all time stamps of sample data
Modify policy of test parameters if any.
end for
4.3.1 Data Preprocessing
The real world databases are highly susceptible to noisy and missing data. So
various preprocessing techniques can be used to improve the quality of data and
thereby improve the prediction results. Data cleaning can be used to
ll in the
missing values. Data transformation can be used to improve the accuracy, speed and
ef
ciency of the algorithms used. Here the data is normalized using the Z-score
normalization where the values of an attribute, A, are normalized based on the mean
ð
Þ
ð r a Þ
v
and standard deviation
of the attribute. The normalized value v of v can be
obtained as:
v 0 ¼ð v v Þ= r a
ð 8 Þ
 
Search WWH ::




Custom Search