Slide 24: at making a direct comparison between the random clustering hyper-heuristic (RCHH) and the adaptive choice-function based clustering hyper-heuristic (ACHH)
The obtained results indicate that although the effects of these methods are very dependent on when and how long they are applied to a solution in the framework, they still have been designed
in an appropriate way to be able to explore different areas of the search space effectively.
Since we use the same low-level heuristics in both frameworks, the significant
improvement of ACHH compared to RCHH is owing to the self-adaptive
phenomenon throughout the hyper-heuristic search. This characteristic takes
appropriate care of the proportion of exploitation/exploration by adjusting parameters
, and
in every iteration. Meanwhile, in RCHH, choosing the
low-level heuristic randomly may lead the solution to the area of the search
space where it is difficult to move quickly to another area. For instance, applying
proposed low-level heuristics which only pay attention to minimising distance,
without taking care of balancing clusters such as Domino, Pair or even Join,
might lead the space to an area with very high quality in terms of overall total
distance and maximum distance but very low quality in relation to balancing. In
this situation, moving the solution space back to a space resulting in a balanced
solution might cause paying too high a penalty in the objective function.
We believe
that this may be caused by the ability of ACHH to learn in the early stages,
which makes the framework more robust and stable, even in small instances.
In other words, adjusting the parameters , and
in an adaptive manner
may speed up the convergence of a hyper-heuristic without knowledge of first
iterations towards a learning framework as iteration increases. 110