Résultats de recherche pour « Branch-and-bound »

Abstract The branch and bound algorithm is mostly used to solve global optimization problems with integer variables (unknowns), linear objective and linear constraints (integer linear programming (LP) problems). The recent increase in the speed of computation on desktop computers has meant that the algorithm can be applied to large problems with several hundred integer variables, which only a decade ago were intractable to any true optimization algorithms. Although a number of algorithms have been proposed for the integer LP problem, the branch and bound technique is used in almost all of the available computer software. This technique has proven to be reasonably efficient for practical problems because it has the embedded advantage of solving continuous linear programmes as subproblems. Furthermore, in some situations the embedded subproblem is a special case of a LP problem (such as assignment and transportation problems), which have more efficient algorithms than the general LP algorithm. Branch and bound methods are also easy to implement on parallel processors. An increasing number of applications are currently forthcoming in this area.

Abstract The branch and bound algorithm is mostly used to solve global optimization problems with integer variables (unknowns), linear objective and linear constraints (integer linear programming (LP) problems). The recent increase in the speed of computation on desktop computers has meant that the algorithm can be applied to large problems with several hundred integer variables, which only a decade ago were intractable to any true optimization algorithms. Although a number of algorithms have been proposed for the integer LP problem, the branch and bound technique is used in almost all of the available computer software. This technique has proven to be reasonably efficient for practical problems because it has the embedded advantage of solving continuous linear programmes as subproblems. Furthermore, in some situations the embedded subproblem is a special case of a LP problem (such as assignment and transportation problems), which have more efficient algorithms than the general LP algorithm. Branch and bound methods are also easy to implement on parallel processors. An increasing number of applications are currently forthcoming in this area.

Abstract Task scheduling in cloud computing is a critical optimization challenge due to resource heterogeneity and dynamic workloads. Metaheuristic algorithms, such as the Flower Pollination Algorithm (FPA), provide fast and near-optimal solutions but may lack precision in complex scenarios. After a comparative evaluation on the GoCJ dataset, FPA was selected for its favorable trade-off between solution quality and execution time. This work introduces HMBB-Sched, a hybrid algorithm that combines FPA with the exact Branch and Bound (B&B) method to enhance solution quality and convergence.To further improve scalability, we propose Parallel_HMBB-Sched, which processes task clusters in parallel using multithreading. All experiments were conducted using CloudSim and the GoCJ dataset, ensuring consistency and reliability of the evaluation. Results show that the hybrid method significantly reduces makespan and execution time compared to standalone FPA. The multithreaded version achieves considerable speedups, particularly under high workloads. This study highlights the effectiveness of combining hybridization and parallelism for efficient and scalable task scheduling in cloud environments.