Group influence based improved firefly algorithm for Design Space Exploration of Datapath resource allocation

摘要

Firefly Algorithm which is a recent addition to the evolutionary algorithms, has shown good performance for many multi-objective optimization problems. In this paper, we propose a novel Firefly algorithm for Design Space Exploration of Datapath resource allocation. The Datapath resource allocation problem is NP-Complete and the design space has vast number of design points. To explore the design space in feasible time, the problem is solved using an improved Firefly algorithm. In particular, meeting the constraints presented by different parameters of interest is evaluated as cost based fitness and then solved. The proposed approach modifies Firefly algorithm on four fronts: 1. A new strategy called Group-Influence based attraction, is used for updating fireflies during evolution; 2. To generate diverse and quality initial population, Opposition Based Learning is incorporated to population initialization; 3. In addition to exploration, in order to refine exploitation, Firefly algorithm is hybridized with Tabu search; 4. Tabu search is updated with Lévy flights for finding nearby solutions. The proposed algorithm is compared with other meta-heuristic algorithms with respect to Quality-of-Results and exploration time. Experimental results show that the proposed algorithm outperforms other existing algorithms for standard benchmark instances.