Benchmark results for the Biobjective Traveling Salesman Problem

 

 

 

This webpage supports the article "Design and analysis of stochastic local search for the multiobjective traveling salesman problem" by L. Paquete and T. Stützle.

 

Short-cut:

 

 

 

Three types of biobjective TSP instances were considered:

The instances are available here

 

Five types of SLS components were tested:

 

 

The experimental methodology followed three steps:

 

 

Notes (see publication for more details):

 

 

 

Euclidean instances

 

Size (n)

100

300

500

Search Strategy:

 

Better Relations:

 

Restart

2phase

Restart

-

0.0%

2phase

0.0%

-

 

 

EAF plots: Restart vs. 2phase

 

Better Relations:

 

Restart

2phase

Restart

-

0.0%

2phase

4.1%

-

 

 

EAF plots: Restart vs. 2phase

 Better Relations:

 

Restart

2phase

Restart

-

0.0%

2phase

10.1%

-

 

 

EAF plots: Restart vs. 2phase

Neighborhood Structure

 

Better Relations:

 

2-opt

3-opt

2-opt

-

0.0%

3-opt

1.2%

-

 

 

EAF plots: 2-opt vs. 3-opt

 

 

Better Relations:

 

2-opt

3-opt

2-opt

-

0.0%

3-opt

43.0%

-

 

 

EAF plots: 2-opt vs. 3-opt

 

 

Better Relations:

 

2-opt

3-opt

2-opt

-

0.0%

3-opt

53.1%

-

 

 

EAF plots: 2-opt vs. 3-opt

 

Component-wise Step

 

Better Relations:

 

False

True

False

-

0.0%

True

0.0%

-

 

 

EAF plots: False vs. True

 

 

 

 

Not tested; always shown to have a positive effect in preliminary experiments.

 

 

 

Not tested; always shown to have a positive effect in preliminary experiments.

Search Length

 

Better Relations:

 

0

50

100

0

-

0.0%

0.0%

50

15.1%

-

0.0%

100

20.4%

0.0%

-

 

 

EAF plots: 0 vs. 50, 50 vs. 100

 

 

Better Relations:

 

0

50

100

0

-

0.0%

0.0%

50

65.3%

-

0.0%

100

80.0%

0.0%

-

 

 

EAF plots: 0 vs. 50, 50 vs. 100

 

 

Better Relations:

 

0

50

100

0

-

0.0%

0.0%

50

71.8%

-

0.0%

100

79.9%

0.0%

-

 

 

EAF plots: 0 vs. 50, 50 vs. 100

 

Number of Scalarizations

 

Better Relations:

 

n

5n

10n

n

-

0.0%

0.0%

5n

0.0%

-

0.0%

10n

0.0%

0.0%

-

 

 

EAF plots: n vs. 5n, 5n vs. 10n

 

 

Better Relations:

 

n

5n

10n

n

-

0.0%

0.0%

5n

0.0%

-

0.0%

10n

0.0%

0.0%

-

 

 

EAF plots: n vs. 5n, 5n vs. 10n

 

 

Better Relations:

 

n

5n

10n

n

-

0.0%

0.0%

5n

0.0%

-

0.0%

10n

0.0%

0.0%

-

 

 

EAF plots: n vs. 5n, 5n vs. 10n

 

 

 

 

Random Instances

 

 

Size (n)

100

300

500

Search Strategy:

 

Better Relations:

 

Restart

2phase

Restart

-

0.0%

2phase

6.4%

-

 

 

EAF plots: Restart vs. 2phase  

 

Better Relations:

 

Restart

2phase

Restart

-

0.0%

2phase

31.6%

-

 

 

EAF plots: Restart vs. 2phase

 Better Relations:

 

Restart

2phase

Restart

-

0.0%

2phase

31.7%

-

 

 

EAF plots: Restart vs. 2phase

Neighborhood Structure

 

Better Relations:

 

2-opt

3-opt

2-opt

-

0.0%

3-opt

50.3%

-

 

 

EAF plots: 2-opt vs. 3-opt

 

 

Better Relations:

 

2-opt

3-opt

2-opt

-

0.0%

3-opt

67.4%

-

 

 

EAF plots: 2-opt vs. 3-opt

 

 

Better Relations:

 

2-opt

3-opt

2-opt

-

0.0%

3-opt

75.0%

-

 

 

EAF plots: 2-opt vs. 3-opt

 

Component-wise Step

 

Better Relations:

 

False

True

False

-

0.0%

True

0.0%

-

 

 

EAF plots: False vs. True  

 

 

Not tested; always shown to have a positive effect in preliminary experiments.

 

 

 

Not tested; always shown to have a positive effect in preliminary experiments.

 

 

Search Length

 

Better Relations:

 

0

50

100

0

-

0.0%

0.0%

50

42.6%

-

0.0%

100

54.9%

0.0%

-

 

 

EAF plots: 0 vs. 50, 50 vs. 100

 

 

Better Relations:

 

0

50

100

0

-

0.0%

0.0%

50

67.0%

-

0.0%

100

77.7%

0.0%

-

 

 

EAF plots: 0 vs. 50, 50 vs. 100

 

 

Better Relations:

 

0

50

100

0

-

0.0%

0.0%

50

69.0%

-

0.0%

100

74.1%

0.0%

-

 

 

EAF plots: 0 vs. 50, 50 vs. 100

 

Number of Scalarizations

 

Better Relations:

 

n

5n

10n

n

-

0.0%

0.0%

5n

0.0%

-

0.0%

10n

0.0%

0.0%

-

 

 

EAF plots: n vs. 5n, 5n vs. 10n

 

 

Better Relations:

 

n

5n

10n

n

-

0.0%

0.0%

5n

8.3%

-

0.0%

10n

11.9%

0.0%

-

 

 

EAF plots: n vs. 5n, 5n vs. 10n

 

 

Better Relations:

 

n

5n

10n

n

-

0.0%

0.0%

5n

7.0%

-

0.0%

10n

10.5%

0.0%

-

 

 

EAF plots: n vs. 5n, 5n vs. 10n

 

 

 

Mixed Instances

 

 

Size (n)

100

300

500

Search Strategy:

 

Better Relations:

 

Restart

2phaseE

2phaseN

Restart

-

0.0%

0.0%

2phaseE

0.0%

-

0.0%

2phaseN

0.0%

0.0%

-

 

 

EAF plots: Restart vs. 2phaseE, Restart vs. 2phaseR, 2phaseE vs. 2phaseR

 

Better Relations:

 

Restart

2phaseE

2phaseN

Restart

-

0.0%

0.0%

2phaseE

5.2%

-

0.0%

2phaseN

2.1%

0.0%

-

 

 

EAF plots: Restart vs. 2phaseE, Restart vs. 2phaseR, 2phaseE vs. 2phaseR

 Better Relations:

 

Restart

2phaseE

2phaseN

Restart

-

0.0%

0.0%

2phaseE

15.8%

-

0.0%

2phaseN

6.9%

0.0%

-

 

 

EAF plots: Restart vs. 2phaseE, Restart vs. 2phaseR, 2phaseE vs. 2phaseR

Neighborhood Structure

 

Better Relations:

 

2-opt

3-opt

2-opt

-

0.0%

3-opt

6.5%

-

 

 

EAF plots: 2-opt vs. 3-opt

 

 

Better Relations:

 

2-opt

3-opt

2-opt

-

0.0%

3-opt

22.3%

-

 

 

EAF plots: 2-opt vs. 3-opt

 

 

Better Relations:

 

2-opt

3-opt

2-opt

-

0.0%

3-opt

26.2%

-

 

 

EAF plots: 2-opt vs. 3-opt  

 

Component-wise Step

 

Better Relations:

 

False

True

False

-

0.0%

True

0.0%

-

 

 

EAF plots: False vs. True

 

 

 

Not tested; always shown to have a positive effect in preliminary experiments.

 

 

 

 

Not tested; always shown to have a positive effect in preliminary experiments.

 

 

Search Length

 

Better Relations:

 

0

50

100

0

-

0.0%

0.0%

50

22.8%

-

0.0%

100

26.6%

0.0%

-

 

 

EAF plots: 0 vs. 50, 50 vs. 100

 

 

Better Relations:

 

0

50

100

0

-

0.0%

0.0%

50

34.6%

-

0.0%

100

52.8%

0.0%

-

 

 

EAF plots: 0 vs. 50, 50 vs. 100

 

 

Better Relations:

 

0

50

100

0

-

0.0%

0.0%

50

40.0%

-

0.0%

100

49.7%

0.0%

-

 

 

EAF plots: 0 vs. 50, 50 vs. 100

 

Number of Scalarizations

 

Better Relations:

 

n

5n

10n

n

-

0.0%

0.0%

5n

0.0%

-

0.0%

10n

0.0%

0.0%

-

 

 

EAF plots: n vs. 5n, 5n vs. 10n

 

 

Better Relations:

 

n

5n

10n

n

-

0.0%

0.0%

5n

0.9%

-

0.0%

10n

7.3%

0.0%

-

 

 

EAF plots: n vs. 5n, 5n vs. 10n

 

 

Better Relations:

 

n

5n

10n

n

-

0.0%

0.0%

5n

1.1%

-

0.0%

10n

7.6%

0.0%

-

 

 

EAF plots: n vs. 5n, 5n vs. 10n

 

 

 

 

Comparison with MOGLS

 

Only the EAF plots are shown.

For the 3 objective instances, we show the EAF plots with parallel coordinates for a given range of differences.

 

 

 

 

Size

 

 

100

300

 

2 Objectives

 

Euclidean

 

Ours vs. MOGLS

 

 

Ours vs. MOGLS

 

 

 

RDM

 

 

Ours vs. MOGLS

 

 

Ours vs. MOGLS

 

3 Objectives

 

 

Euclidean

 

 

(80%,100%]: Ours vs. MOGLS

(60%,80%]: Ours vs. MOGLS

 

 

(80%, 100%]: Ours vs. MOGLS

(60%,80%]: Ours vs. MOGLS

 

 

 

RDM

 

 

(80%, 100%]: Ours vs. MOGLS

(60%,80%]: Ours vs. MOGLS

(40%,60%]: Ours vs. MOGLS

 

 

(80%, 100%]: Ours vs. MOGLS

(60%, 80%]: Ours vs. MOGLS