README.md

Sensitivity Analysis of Weight and Threshold with Respect to Experimental Results

The following tables show the effects of threshold/weight on different datasets. The results show that the performance are not so sensitive to the threshold/weight.

Table 1：Experimental Results under Different Hyperparameters on the PHD Dataset

${W_{1}}$	${W_{2}}$	${W_{3}}$	${W_{4}}$	${T}$	$\boldsymbol{F_{1}}$	Acc
0.25	0.1	0.15	0.5	$\mu$	53.8	67.3
0.5	0.2	0.1	0.2	$\mu$	54.5	67.6
0.7	0.1	0.1	0.1	$\mu$	55.4	68.3
0.25	0.5	0.15	0.1	$\mu+2\sigma$	52.7	65.6
0.5	0.1	0.3	0.1	$\mu+2\sigma$	53.1	66.2
0.7	0.15	0.1	0.05	$\mu+2\sigma$	54.0	67.1

Table 2：Experimental Results under Different Hyperparameters on the WikiBio Dataset

${W_{1}}$	${W_{2}}$	${W_{3}}$	${W_{4}}$	${T}$	$\boldsymbol{F_{1}}$	Acc
0.25	0.5	0.15	0.1	$\mu+\sigma$	88.6	79.8
0.5	0.2	0.15	0.15	$\mu+\sigma$	89.3	80.3
0.7	0.1	0.1	0.1	$\mu+\sigma$	89.7	81.0
0.25	0.3	0.35	0.1	$\mu+2\sigma$	89.1	81.7
0.5	0.1	0.3	0.1	$\mu+2\sigma$	89.6	82.5
0.7	0.15	0.1	0.05	$\mu+2\sigma$	90.7	83.2

Results on both Table 1 and Table 2 show that under different hyperparameter (weight/threshold) settings, the variation in experimental results is not significant and remains relatively stable.

Table 3：Analysis of One-Way ANOVA Results between Different Hyperparameters and Metric Values on the PHD Dataset

	${W_{1}}$	${W_{2}}$	${W_{3}}$	${W_{4}}$	${T}$
$\boldsymbol{F_{1}}$	0.06	0.12	0.09	0.85	0.17
Acc	0.07	0.15	0.09	0.18	0.23

Table 4: Analysis of One-Way ANOVA Results between Different Hyperparameters and Metric Values on the WikiBio Dataset

	${W_{1}}$	${W_{2}}$	${W_{3}}$	${W_{4}}$	${T}$
$\boldsymbol{F_{1}}$	0.08	0.12	0.08	0.04	0.33
Acc	0.01	0.04	0.02	0.09	0.27

One-way ANOVA results on Table 3 and Table 4 show that all p-values are greater than 0.1, indicating that the experimental results are not sensitive to the hyperparameter settings