Information Technology Reference
In-Depth Information
13738.4
LAHC
sLAHC
12238.4
10738.4
9238.44
1
3000.8
6000.6
9000.4
12000.2
Iterations
Fig. 3.
Iterations-to-target evaluation (LAHC/sLAHC). LAHC and sLAHC have been
applied to 25 instances each with 15000 iterations in all cases.
5Con lu on
We have applied Late Acceptance Hill Climbing to the Traveling Purchaser Prob-
lem. We combined several ideas from literature for building initial solutions and
for defining neighborhoods. We tested our approach on academic benchmarks
and the results indicate that the LAHC procedure is suitable for solving the
TPP. Furthermore, we also presented a simplified application of LAHC which
led to promising results, especially for instances with a smaller number of prod-
ucts. The relatively weak performance of LAHC for small instances seems to in-
dicate a small drawback of LAHC in comparison to other metaheuristics. While
LAHC requires less effort for parameter setting, it does not allow a fine tuning
of acceptance probabilities like, e.g., simulated annealing allows. On the other
hand, LAHC might work well with an auto-adaptive parameter tuning that ad-
justs the neighborhood construction methods and the LAHC fitness array length.
This could be a promising future research approach.
