Learning is an important part of the immune system by which the immune system maintains a memory of the infections it has encountered to protect against future attacks. In this paper we incorporate the mechanism of maintaining a memory in the recently proposed GC-AIS algorithm. GC-AIS has shown good performance on the static set cover problem (SCP) in recent work  and we are interested in investigating the merits of GC-AIS in a dynamic setting. We compare the affect of GC-AIS with and without a memory approach on the dynamic SCP instances, which are created with varying degrees of modifications to instances from . Three types of modifications are proposed in the paper by adding, removing or editing the subsets from the original problem instances. It is shown that for the case of adding subsets to the original instance using our memory approach is always beneficial while for the case of removing subsets using our memory approach almost always results in worse performance than when not utilising memory. Finally in the cases with editing subsets it is shown that for lower levels of modification using our memory approach gives better results while when the level of modification is higher our memory based approach is worse than using no memory.
|Title of host publication||MIC 2015: The XI Metaheuristics International Conference|
|Publisher||University of Lille 1|
|Publication status||Published - 2015|
|Event||11th Metaheuristics Internation Conference (MIC 2015) - Agadir, Morocco|
Duration: 07 Jun 2015 → 10 Jun 2015
|Conference||11th Metaheuristics Internation Conference (MIC 2015)|
|Period||07 Jun 2015 → 10 Jun 2015|