Table III. The algorithmic parameter...

Table III.

The algorithmic parameter tuning results based on the full factorial design methodology

Algorithm	Parameter	Candidate values	Selected value
UIMA	Population size (PopSize) ^a	[240; 280; 320]	240
UIMA	Migration criterion (λ)	[0.1; 0.2; 0.3]	0.1
UIMA	Number of immigrants (N^img)	[3; 4; 5]	4
UIMA	Stopping criterion (MaxIter)	[1200; 1400; 1600]	1600
EA	Population size (PopSize)^b	[40; 50; 60]	60
EA	Crossover probability (σ^c)	[0.3; 0.5; 0.8]	0.8
EA	Mutation probability (σ^m)	[0.01; 0.04; 0.06]	0.06
EA	Number of chromosomes attending each tournament (TourSize)	[20; 30; 40]	40
EA	Stopping criterion (MaxGen)^b	[4000; 5000; 6000]	6000
PSO	Population size (PopSize)^b	[40; 50; 60]	60
PSO	Cognition component (C₁)	[1.5; 2.0; 2.5]	2.0
PSO	Social component (C₂)	[1.5; 2.0; 2.5]	1.5
PSO	Inertia weight (W)	[0.3; 0.5; 0.8]	0.5
PSO	Stopping criterion (MaxIter)^b	[4000; 5000; 6000]	6000
EDA	Population size (PopSize)^b	[40; 50; 60]	60
EDA	Shaking coefficient (ε)	[0.1; 0.3; 0.5]	0.1
EDA	Elitism coefficient (ψ)	[0.4; 0.6; 0.8]	0.6
EDA	Stopping criterion (MaxGen)^b	[4000; 5000; 6000]	6000
DE	Population size (PopSize)^b	[40; 50; 60]	60
DE	Mutation coefficient (α)	[0.4; 0.6; 0.8]	0.8
DE	Crossover probability (σ^c)	[0.3; 0.5; 0.6]	0.3
DE	Stopping criterion (MaxGen)^b	[4000; 5000; 6000]	6000
VNS	Local search size $(S_{l s}^{V N S})$	[15; 20; 25]	20
VNS	Stopping criterion (MaxIter)	[4000; 5000; 6000]	6000
TS	Tabu list size $(S_{t l}^{T S})$	[10; 15; 20]	10
TS	Local search size $(S_{l s}^{T S})$	[15; 20; 25]	20
TS	Stopping criterion (MaxIter)	[4000; 5000; 6000]	6000
SA	Initial Boltzmann temperature (τ)	[2000; 3000; 3500]	3500
SA	Temperature interval (Δτ)	[0.10; 0.30; 0.50]	0.50
SA	Stopping criterion (MaxIter)	[4000; 5000; 6000]	6000

Algorithm	Parameter	Candidate values	Selected value
UIMA	Population size (PopSize) ^a	[240; 280; 320]	240
UIMA	Migration criterion (λ)	[0.1; 0.2; 0.3]	0.1
UIMA	Number of immigrants (N^img)	[3; 4; 5]	4
UIMA	Stopping criterion (MaxIter)	[1200; 1400; 1600]	1600
EA	Population size (PopSize)^b	[40; 50; 60]	60
EA	Crossover probability (σ^c)	[0.3; 0.5; 0.8]	0.8
EA	Mutation probability (σ^m)	[0.01; 0.04; 0.06]	0.06
EA	Number of chromosomes attending each tournament (TourSize)	[20; 30; 40]	40
EA	Stopping criterion (MaxGen)^b	[4000; 5000; 6000]	6000
PSO	Population size (PopSize)^b	[40; 50; 60]	60
PSO	Cognition component (C₁)	[1.5; 2.0; 2.5]	2.0
PSO	Social component (C₂)	[1.5; 2.0; 2.5]	1.5
PSO	Inertia weight (W)	[0.3; 0.5; 0.8]	0.5
PSO	Stopping criterion (MaxIter)^b	[4000; 5000; 6000]	6000
EDA	Population size (PopSize)^b	[40; 50; 60]	60
EDA	Shaking coefficient (ε)	[0.1; 0.3; 0.5]	0.1
EDA	Elitism coefficient (ψ)	[0.4; 0.6; 0.8]	0.6
EDA	Stopping criterion (MaxGen)^b	[4000; 5000; 6000]	6000
DE	Population size (PopSize)^b	[40; 50; 60]	60
DE	Mutation coefficient (α)	[0.4; 0.6; 0.8]	0.8
DE	Crossover probability (σ^c)	[0.3; 0.5; 0.6]	0.3
DE	Stopping criterion (MaxGen)^b	[4000; 5000; 6000]	6000
VNS	Local search size $(S_{l s}^{V N S})$	[15; 20; 25]	20
VNS	Stopping criterion (MaxIter)	[4000; 5000; 6000]	6000
TS	Tabu list size $(S_{t l}^{T S})$	[10; 15; 20]	10
TS	Local search size $(S_{l s}^{T S})$	[15; 20; 25]	20
TS	Stopping criterion (MaxIter)	[4000; 5000; 6000]	6000
SA	Initial Boltzmann temperature (τ)	[2000; 3000; 3500]	3500
SA	Temperature interval (Δτ)	[0.10; 0.30; 0.50]	0.50
SA	Stopping criterion (MaxIter)	[4000; 5000; 6000]	6000

Note:

^aThe UIMA population was evenly distributed between its four islands (i.e. the UIMA population of 240 individuals was distributed between its four islands in the way that the EA, PSO, EDA, and DE sub-populations would have 60 individuals per sub-population); ^bthe population size and stopping criterion that were set for EA, PSO, EDA, and DE when they were executed independently

[ViewLarge]

Sharing Unavailable