Although it is well-known that a proper balancing between exploration and exploitation plays a central role on the performances of any evolutionary algorithm, what instead becomes crucial for both is the life time with which any offspring maturate and learn. Setting an appropriate lifespan helps the algorithm in a more efficient search as well as in fruitful exploitation of the learning discovered. Thus, in this research work we present an experimental study conducted on eleven different age assignment types, and performed on a classical genetic algorithm, with the aim to (i) understand which one provides the best performances in term of overall efficiency, and robustness; (ii) produce an efficiency ranking; and, (iii) as the most important goal, verify and prove if the tops, or most, or the whole ranking previously produced on an immune algorithm coincide with that produced for genetic algorithm. From the analysis of the achievements obtained it is possible to assert how the two efficiency rankings are roughly the same, primarily for the top 4 ranks. This also implies that the worst option obtained for the immue algorithm continues to be a bad choice even for the genetic algorithm. The most important outcomes that emerge from this research work are respectively (1) the age assignment to be avoided, from which are obtained bad performances; and (2) a reliable age to be assigned to any offspring for having, with high probability, robust and efficient performances.
|Titolo:||Optimizing the Individuals Maturation for Maximizing the Evolutionary Learning|
PAVONE, MARIO FRANCESCO (Corresponding)
|Data di pubblicazione:||2017|
|Appare nelle tipologie:||4.2 Abstract in Atti di convegno|