ACLib2 Flowshop Study

Build experiments

Problem space:

• Instance attributes:

  • Number of jobs: 50, 100, 200
  • Number of machines: 20, 40
  • Processing time distributions: uniform, erlang, exponential
  • Processing time correlations: random, job-correlated, machine-correlated

• Model attributes:

  • Constraints: PERM, NOWAIT, NOIDLE
  • Objectives: FLOWTIME, MAKESPAN
  • Budgets: low, high

Comparing all performances with respect to each problem (blocks):

## 
##  Friedman rank sum test
## 
## data:  y, groups and blocks
## Friedman chi-squared = 24.274, df = 3, p-value = 2.19e-05

Comparing the performances with respect to the default NEH:

##          DEFAULT
## IRACE3   0.00078
## PARAMILS 0.00108
## SMAC3    0.00052

Comparing the performance of the different tuners:

## 
##  Friedman rank sum test
## 
## data:  y, groups and blocks
## Friedman chi-squared = 0.74286, df = 2, p-value = 0.6897

## 
##  Friedman rank sum test
## 
## data:  y, groups and blocks
## Friedman chi-squared = 0.74286, df = 2, p-value = 0.6897

Statistics of the performances:

tunner	min	q25	q50	q75	max
IRACE3	0	0	0.0026163	0.6684956	2.990654
PARAMILS	0	0	0.0000000	0.3649868	5.715786
SMAC3	0	0	0.0000000	0.6140526	3.919670