Slide 1
Slide 1 text
Automatic Design of Ant Algorithms
with Grammatical Evolution
Jorge Tavares
Francisco B. Pereira
CISUC, University of Coimbra, Portugal
Slide 2
Slide 2 text
how do we design
an Ant algorithm?
Slide 3
Slide 3 text
“A man provided with paper, pencil, and rubber,
and subject to strict discipline, is in effect a
universal machine.”
Alan Turing
Slide 4
Slide 4 text
how can we
automatically design
an Ant algorithm?
Slide 5
Slide 5 text
Grammatical Evolution + Ant Structures
= Evolutionary Ant Algorithms
Slide 6
Slide 6 text
running the system
Initial Population
Access Fitness of
Population
Satisfactory
Individual exist ?
Select Individuals
Make Random
Changes
Return Best
Individual
Ant
Algorithms
Run Ant Algorithm
to solve a problem
Best Ant
Algorithm for
the problem
Slide 7
Slide 7 text
::= (aco )
::= (repeat-until
)
::= (foreach-ant make-solution-with
(if (local-update-trails )))
::= (roulette-selection)
| (q-selection )
| (random-selection)
::= (progn )
...
...
Slide 8
Slide 8 text
the system is able
to generate all
main Ant algorithms
Ant System (AS), Elitist Ant System (EAS), Rank-
Based Ant System (RAS), Ant Colony System (ACS),
Max-Min Ant System (MMAS)
Slide 9
Slide 9 text
the search space contains
many other combinations
that define
alternative Ant algorithms
Slide 10
Slide 10 text
does search converge
to manually designed
Ant algorithms?
Slide 11
Slide 11 text
can we find novel Ant
algorithms?
Slide 12
Slide 12 text
test problem:
traveling salesman problem
selected instances from TSPLIB:
learning: eil76, pr76, gr96
testing: att48, eil51, berlin52, kroA100, lin105, gr137, u159, d198, pr226, lin318
Slide 13
Slide 13 text
0!
0.05!
0.1!
0.15!
0.2!
0.25!
0.3!
0.35!
1! 5! 9! 13! 17! 21! 25!
Distance to optimum!
Generations!
median of best individuals
Slide 14
Slide 14 text
evolution summary (30 runs)
Slide 15
Slide 15 text
search does not converge
to manually designed
architectures
Slide 16
Slide 16 text
main components frequency
Slide 17
Slide 17 text
evolved example: ei7618
Slide 18
Slide 18 text
but... are the evolved
Ant algorithms effective?
Slide 19
Slide 19 text
optimization results
Slide 20
Slide 20 text
statistical analysis
Slide 21
Slide 21 text
comparison with standard algorithms
Slide 22
Slide 22 text
statistical analysis
Slide 23
Slide 23 text
hybrid between the two best evolved
Slide 24
Slide 24 text
1) evolution is able to discover original
architectures
2) best evolved strategies exhibit a good
generalization capability
3) competitive with human-designed variants
4) hybrid architecture based on evolved
strategies is effective
conclusions
Slide 25
Slide 25 text
“We can only see a short distance
ahead, but we can see plenty
there that needs to be done.”
Alan Turing
Slide 26
Slide 26 text
Thank you. Questions?
Special thanks to Alex Wild for all the Ant pictures @alexanderwild.com