TABLE X.

Parameter Trial Ranges of All Algorithms and Recommended Optimal Settings of TI-/TII-APJCSC on Real-Life Data Sets

Datasets	Trial ranges (σ)	Recommended optimal settings
		TI-APJCSC	TII-APJCSC
Banana	[0.015:0.002:0.085]	σ = 0.043, η = 0.5	σ = 0.037, η = 0.3
Wisconsin	[0.05:0.02:0.85]	σ = 0.11, η = 0.3	σ = 0.11, η = 0.1
Led7digit	[0.277:0.02:1.077]	σ ∈ = [0.517,0.737], η ∈ [0.7,0.9]	σ ∈ [0.697,0.957], η ∈ [0.3,0.4]
Wine	[0.05:0.02:1.55]	σ ∈ [0.25,0.29], η ∈ [0.8,0.9]	σ ∈ [0.27,0.31], η ∈ [0.5,0.8]
Waveform-21	[0.3:0.02:1.3]	σ ∈ [0.34,0.74], η ∈ [0.2,0.9]	σ ∈ [0.92,1.1], η ∈ [0.8,0.9]
USPS-3568	[0.3:0.02:2.2]	σ ∈ [1.12,1.3], η ∈ [0.6,0.7]	σ ∈ [0.5,0.72], η = 0.9
JAFFE	[0.5:0.035:5.1]	σ ∈ [2.67,3.3], η ∈ [0.2,0.7]	σ ∈ [2.32,3.23], η ∈ [0.2,0.9]
20news	[0.3:0.05:2.3]	σ ∈ [0.45,0.55], η ∈ [0.1,0.3]	σ ∈ [0.35,0.45], η ∈ [0.1,0.3]
Berke-296059	[0.01:0.0012:0.08]	σ = 0.0220, η = 0.8	σ = 0.0484, η = 0.3

Note: Each interval or specific value of optimal settings is achieved by 10 times of implementations of TI-/TII-APJCSC on the same dataset but with different supervision (Except for the three KEEL datasets, where the supervision is invariant). If the ten results are inconsistent, the referenced interval is listed; otherwise, the specific value is given.