Modified Selection and Search in Learning Automata Based Artificial Bee Colony in Noisy Environment
Pratyusha Rakshit, Amit Konar, Atulya K. Nagar · 2019
The paper proposes two novel extensions of the learning automata induced noisy bee colony (LANBC) to handle the presence of stochastic noise in the fitness landscape. The reward/penalty based reinforcement learning scheme of the stochastic learning automata (SLA) of LANBC helps a solution to prudently select a sample size for its periodic fitness evaluations based on the fitness variance in its local neighborhood. However, the LANBC suffers from two stalemates, including weak search dynamic in the presence of noise and the deterministic parental selection scheme leading to dismissal of quality solutions and loss of population diversity. The paper overcomes these impasses by employing two novel strategies. First, the search dynamic is amended with an aim to generate promising offspring solutions by appropriately tuning the control parameter based on the contamination effect of noise on other population members. Second, a modified probabilistic crowding induced niching behavior is introduced to promote quality solutions to the next generation to ensure both the population quality and the population diversity. Computer simulations undertaken on the noisy versions of a set of 28 benchmark functions reveal that the proposed algorithm outperforms its contenders with respect to function error value.