By Jérémie
DuboisLacoste, Manuel LópezIbáñez and Thomas Stützle
January 2010 (last update: January 2010)
Table of Contents 
Jérémie DuboisLacoste, Manuel
LópezIbáñez, and Thomas Stützle. Adaptive “Anytime”
TwoPhase Local Search. In Learning and Intelligent Optimization,
4th International Conference, LION 4, volume 6073 of Lecture Notes in
Computer Science, pages 52–67. Springer, Heidelberg, Germany,
2010.
★ Best paper award
[ Bibtex ] [ doi: 10.1007/9783642138003_5 ]
TwoPhase Local Search (TPLS) is a general algorithmic framework for
multiobjective optimization. TPLS transforms the multiobjective problem into
a sequence of singleobjective ones by means of weighted sum aggregations.
This paper studies different sequences of weights for defining the aggregated
problems for the biobjective case.
In particular, we propose two weight setting strategies that show better
anytime search characteristics than the original weight setting strategy used
in the TPLS algorithm.
Keywords: Metaheuristics, TwoPhase Local Search, Anytime
property, Adaptive strategy, Scheduling, Flowshop
These plots show the development of the hypervolume (mean over 15
runs) over the number of scalarizations, for Regular Anytime (RATPLS) against the classical strategies. 
Makespan and Sum of flowtimes



50x20_1

50x20_2

50x20_3

50x20_4

50x20_5

100x20_1

100x20_2

100x20_3

100x20_4

100x20_5

Makespan and Total Tardiness


50x20_1

50x20_2

50x20_3

50x20_4

50x20_5

100x20_1

100x20_2

100x20_3

100x20_4

100x20_5

These plots show the development of the hypervolume (mean over 15 runs) over the number of scalarizations, for the Anytime strategies (RATPLS, ANTPLS and AFTPLS) against Double (DTPLS). 
Makespan and Sum of flowtimes



50x20_1

50x20_2

50x20_3

50x20_4

50x20_5

100x20_1

100x20_2

100x20_3

100x20_4

100x20_5

Makespan and Total Tardiness


50x20_1

50x20_2

50x20_3

50x20_4

50x20_5

100x20_1

100x20_2

100x20_3

100x20_4

100x20_5

These plots show the difference of empirical attainment functions (over 15 runs) for Regular Anytime and Adaptive Anytime. 
Makespan and Sum of flowtimes


50x20_1 
50x20_2 
50x20_3 
50x20_4 
50x20_5 
100x20_1 
100x20_2 
100x20_3 
100x20_4 
100x20_5 
Makespan and Total Tardiness

50x20_1 
50x20_2 
50x20_3 
50x20_4 
50x20_5 
100x20_1 
100x20_2 
100x20_3 
100x20_4 
100x20_5 
These plots show the difference of empirical attainment functions (over 15 runs) for Adaptive Anytime and Double TPLS. 
Makespan and Sum of flowtimes


50x20_1 
50x20_2 
50x20_3 
50x20_4 
50x20_5 
100x20_1 
100x20_2 
100x20_3 
100x20_4 
100x20_5 
Makespan and Total Tardiness

50x20_1 
50x20_2 
50x20_3 
50x20_4 
50x20_5 
100x20_1 
100x20_2 
100x20_3 
100x20_4 
100x20_5 