IEEE Computational Intelligence Magazine - May 2022 - 79

for the cheap task; 3) :m the penalty
parameter. The values of other parameters
are fixed to analyze one parameter visually.
Figure 5 shows the average AUC
value of EMTAUC versus varying d on
diabetes and australian datasets. The
average AUC value of EMTAUC varies
with d on all datasets as shown in Fig. S3
(see Supplementary Material). The value
of d is set from 5 to 40 in steps of 5. It can
be seen that the value of d has an essential
effect on the performance of
EMTAUC. The parameter affects the
frequency of dynamic adjustment of the
cheap task. Moreover, in Fig. 5(a), the
value of d should be set to 5 to obtain
the best average objective value on the
diabetes dataset. However, in Fig. 5(b),
EMTAUC can obtain the best performance
on the australian dataset if
d 10 .=
Thus, the collection of the parameter on
the diabetes dataset may not be suitable
for the others. d should be set to a relatively
small value. In addition, the change
curve is located above the straight line
corresponding to the average AUC value
obtained by Single-task GA in most
cases, showing that no matter what the
value of d is, useful knowledge will be
transferred across tasks to improve the
convergence performance.
Figure 6 shows the average objective
value of EMTAUC versus varying s on
diabetes, fourclass, german, and splice
datasets. The average objective value of
EMTAUC varies with s on all datasets as
shown in Fig. S4 (see Supplementary
Material). In Fig. 6, the average objective
value increases with increasing
computational costs. Figure 6 (a) shows
that the performance of EMTAUC on
the diabetes dataset is the best in all cases
when s is set to 0.2. Figure 6 (b) shows
that when s is set to 0.3, the performance
of EMTAUC is the best in all
cases. With the increase of s, the AUC
obtained by EMTAUC first increases
and then decreases. This phenomenon
can be found in Fig. S4 (a), (b), (d), (e),
(f), (g), (h), (k), (l), (m), (n), and (o). This
phenomenon appears because the relatively
high ratio of the whole dataset
may own the better representation of
the original dataset, but a higher value
of s may break the balance of the
MAY 2022 | IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE 79
TABLE VI The comparison of EMTAUC with and without the dynamic adjustment strategy of AUCc (EMTAUC and EMTAUC-I) in terms of AUC.
DATASET
diabetes
EMTAUC-MFEAII
8.26E−01(3.34E−03) .
fourclass
german_scale
splice
usps
australian
a9a
sonar
svmguide1
svmguide3
segment
ijcnn1
satimage
vowel
poker
w/t/l
8.34E−01(4.73E−04)+
7.84E−01(1.76E−03)+
8.19E−01(3.08E−03)+
9.50E−01(1.53E−03) .
9.22E−01(2.60E−03) .
8.90E−01(9.87E−04) .
8.44E−01(8.33E−03)+
9.82E−01(4.41E−03) .
7.42E−01(9.48E−03)+
9.70E−01(2.77E−03) .
8.30E−01(9.62E−03)+
9.97E−01(8.26E−04) .
7.99E−01(3.65E−03) .
5.21E−01(1.01E−03)+
7/8/0
EMTAUC-I-MFEAII
8.21E−01(4.11E−03)
8.25E−01(1.25E−03)
7.78E−01(1.49E−03)
8.11E−01(3.40E−03)
9.49E−01(1.14E−03)
9.25E−01(1.33E−03)
8.88E−01(6.57E−04)
8.21E−01(1.03E−02)
9.84E−01(3.22E−03)
7.33E−01(5.75E−03)
9.72E−01(6.18E−03)
8.18E−01(8.02E−03)
9.97E−01(1.37E−03)
7.98E−01(3.85E−03)
5.07E−01(1.24E−03)
−
EMTAUC-SBGA
8.21E−01(3.87E−03) .
8.29E−01(2.84E−03) .
7.71E−01(4.14E−03)+
7.87E−01(8.48E−03)+
9.37E−01(1.32E−03)+
9.23E−01(1.82E−03) .
8.79E−01(1.68E−03) .
8.40E−01(7.13E−03)+
9.83E−01(4.78E−03) .
7.31E−01(1.12E−02)+
9.66E−01(4.76E−03)+
7.75E−01(1.00E−02)−
9.97E−01(8.28E−04) .
8.01E−01(5.04E−03)+
5.11E−01(3.03E−03)+
8/6/1
EMTAUC-I-SBGA
8.24E−01(4.38E−03)
8.29E−01(2.35E−03)
7.67E−01(3.90E−03)
7.77E−01(3.67E−03)
9.31E−01(2.61E−03)
9.24E−01(1.62E−03)
8.78E−01(1.04E−03)
8.13E−01(1.30E−02)
9.86E−01(5.48E−03)
7.20E−01(9.28E−03)
9.56E−01(6.49E−03)
7.91E−01(1.02E−02)
9.97E−01(7.80E−04)
7.81E−01(3.10E−03)
5.08E−01(1.67E−03)
−
EMTAUC-EMEA
EMTAUC-I-EMEA
8.17E−01(5.50E−03) .
8.30E−01(1.09E−03) .
7.72E−01(6.05E−03)+
7.82E−01(6.92E−03)+
9.31E−01(3.01E−03) .
9.21E−01(1.68E−03) .
8.72E−01(1.25E−03) .
8.39E−01(1.43E−02)+
9.81E−01(3.65E−03) .
7.27E−01(7.76E−03)+
9.58E−01(5.90E−03) .
7.80E−01(8.59E−03)+
9.97E−01(6.23E−04) .
8.01E−01(6.33E−03)+
5.11E−01(2.28E−03)+
7/8/0
8.19E−01(1.44E−02)
8.29E−01(2.12E−03)
7.67E−01(6.53E−03)
7.72E−01(3.58E−03)
9.31E−01(2.61E−03)
9.23E−01(2.38E−03)
8.72E−01(2.44E−03)
8.27E−01(1.08E−02)
9.83E−01(6.16E−03)
7.14E−01(8.02E−03)
9.60E−01(4.77E−03)
7.73E−01(7.34E−03)
9.97E−01(9.94E−04)
7.81E−01(5.61E−03)
5.08E−01(2.08E−03)
IEEE Computational Intelligence Magazine - May 2022

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - May 2022
