A self-adaptive large neighborhood search method for scheduling n jobs on m non-identical parallel machines with mul- tiple time windows is presented. The problems＇ another feature lies in oversubscription, namely no...A self-adaptive large neighborhood search method for scheduling n jobs on m non-identical parallel machines with mul- tiple time windows is presented. The problems＇ another feature lies in oversubscription, namely not all jobs can be scheduled within specified scheduling horizons due to the limited machine capacity. The objective is thus to maximize the overall profits of processed jobs while respecting machine constraints. A first-in- first-out heuristic is applied to find an initial solution, and then a large neighborhood search procedure is employed to relax and re- optimize cumbersome solutions. A machine learning mechanism is also introduced to converge on the most efficient neighborhoods for the problem. Extensive computational results are presented based on data from an application involving the daily observation scheduling of a fleet of earth observing satellites. The method rapidly solves most problem instances to optimal or near optimal and shows a robust performance in sensitive analysis.展开更多
图是一种非常重要的数据结构形式,被广泛用于社交网络、交通网络和搜索引擎等领域。随着图数据规模爆发式增长,存储容量受限,分布式图计算成为处理大规模图数据的焦点。宽度优先搜索(breadth first search,BFS)算法是图遍历和许多图分...图是一种非常重要的数据结构形式,被广泛用于社交网络、交通网络和搜索引擎等领域。随着图数据规模爆发式增长,存储容量受限,分布式图计算成为处理大规模图数据的焦点。宽度优先搜索(breadth first search,BFS)算法是图遍历和许多图分析算法的基础,而在分布式图计算过程中存在严重的通信开销。针对上述问题,本文提出了一种综合的数据压缩编码优化方案,结合位图和变长压缩数组,通过更高的压缩率来降低数据通信开销;此外,还提出了一种点对点异步环形通信策略,进一步降低分布式图计算中计算-通信的同步开销。通过这些优化手段,本文在8节点的分布式集群上对优化后BFS算法的性能进行了系统评估,结果表明,当图数据规模为28时,优化后的BFS算法平均性能为46.79亿条边每秒遍历(giga-traversed edges per second,GTEPS),性能比优化前提升了接近7.82%。展开更多
In this paper, it is supposed that the B&B algorithm finds the first optimal solution after h nodes have been expanded and m active nodes have been created in the state-space tree. Then the lower bound Ω(m+h log ...In this paper, it is supposed that the B&B algorithm finds the first optimal solution after h nodes have been expanded and m active nodes have been created in the state-space tree. Then the lower bound Ω(m+h log h) of the running time for the general sequential B&B algorithm and the lower bound Ω(m/p+h log p) for the general parallel best-first B&B algorithm in PRAM-CREW are proposed, where p is the number of processors available. Moreover, the lower bound Ω(M/p+H+(H/p) log (H/p)) is presented for the parallel algorithms on distributed memory system, where M and H represent total number of the active nodes and that of the expanded nodes processed by p processors, respectively. In addition, a nearly fastest general parallel best-first B&B algorithm is put forward. The parallel algorithm is the fastest one as p = max{hε, r}, where ε = 1/ rootlogh, and r is the largest branch number of the nodes in the state-space tree.展开更多
针对蒙特卡洛树搜索算法(Monte Carlo tree search,MCTS)收敛速度过慢,且在博弈过程中关键节点会出现信息丢失等问题,以中国象棋为载体,构建适用于中国象棋博弈系统的策略价值网络,提出了一种基于统计数据的并行蒙特卡洛树搜索算法(para...针对蒙特卡洛树搜索算法(Monte Carlo tree search,MCTS)收敛速度过慢,且在博弈过程中关键节点会出现信息丢失等问题,以中国象棋为载体,构建适用于中国象棋博弈系统的策略价值网络,提出了一种基于统计数据的并行蒙特卡洛树搜索算法(parallel Monte Carlo tree search based on statistics,SPMCTS)。将并行化的重点设置在MCTS四个步骤中最耗时的扩展和模拟步骤,有效避免了算法执行过程中的等待时差。并且引入一组新统计数据,这些数据用于在MCTS的选择步骤中修改节点的选择策略,保证在进行节点选择时获取和利用更多的可用信息,缓解信息丢失对精度造成的影响。实验结果表明,与现有并行蒙特卡洛树算法相比,SPMCTS在搜索速度上加快了约34%,且在对弈实验中,博弈胜率也能保持在80%左右。验证了SPMCTS的有效性。展开更多
在静态栅格地图中,针对传统蚁群算法进行AGV(Automated Guided Vehicle,自动引导车)路径规划收敛慢且搜索结果容易陷入局部最优的问题,提出一种融合跳点搜索(Jump Point Search,JPS)和双向并行蚁群搜索的改进算法.首先,对实际研究环境...在静态栅格地图中,针对传统蚁群算法进行AGV(Automated Guided Vehicle,自动引导车)路径规划收敛慢且搜索结果容易陷入局部最优的问题,提出一种融合跳点搜索(Jump Point Search,JPS)和双向并行蚁群搜索的改进算法.首先,对实际研究环境进行栅格化建模,使用改进的跳点搜索算法生成双向搜索的初始次优路径,为双向蚁群搜索提供初始搜索方向参考.其次,在双向并行蚁群搜索过程中采用改进的转移概率启发函数,该函数在确定下一个转移节点时考虑了避免AGV与障碍物碰撞的因素,同时通过设计信息素共享机制并结合改进的信息素增量及浓度两种融合模型,共享和更新全局信息素浓度,以更好地探索和优化路径,保证双向路径连结.最后,与传统蚁群算法进行实验结果对比,验证了改进算法的全局搜索能力、效率和安全性.展开更多
基金supported by the National Natural Science Foundation of China (7060103570801062)
文摘A self-adaptive large neighborhood search method for scheduling n jobs on m non-identical parallel machines with mul- tiple time windows is presented. The problems＇ another feature lies in oversubscription, namely not all jobs can be scheduled within specified scheduling horizons due to the limited machine capacity. The objective is thus to maximize the overall profits of processed jobs while respecting machine constraints. A first-in- first-out heuristic is applied to find an initial solution, and then a large neighborhood search procedure is employed to relax and re- optimize cumbersome solutions. A machine learning mechanism is also introduced to converge on the most efficient neighborhoods for the problem. Extensive computational results are presented based on data from an application involving the daily observation scheduling of a fleet of earth observing satellites. The method rapidly solves most problem instances to optimal or near optimal and shows a robust performance in sensitive analysis.
文摘图是一种非常重要的数据结构形式,被广泛用于社交网络、交通网络和搜索引擎等领域。随着图数据规模爆发式增长,存储容量受限,分布式图计算成为处理大规模图数据的焦点。宽度优先搜索(breadth first search,BFS)算法是图遍历和许多图分析算法的基础,而在分布式图计算过程中存在严重的通信开销。针对上述问题,本文提出了一种综合的数据压缩编码优化方案,结合位图和变长压缩数组,通过更高的压缩率来降低数据通信开销;此外,还提出了一种点对点异步环形通信策略,进一步降低分布式图计算中计算-通信的同步开销。通过这些优化手段,本文在8节点的分布式集群上对优化后BFS算法的性能进行了系统评估,结果表明,当图数据规模为28时,优化后的BFS算法平均性能为46.79亿条边每秒遍历(giga-traversed edges per second,GTEPS),性能比优化前提升了接近7.82%。
基金This paper was supported by Ph. D. Foundation of State Education Commission of China.
文摘In this paper, it is supposed that the B&B algorithm finds the first optimal solution after h nodes have been expanded and m active nodes have been created in the state-space tree. Then the lower bound Ω(m+h log h) of the running time for the general sequential B&B algorithm and the lower bound Ω(m/p+h log p) for the general parallel best-first B&B algorithm in PRAM-CREW are proposed, where p is the number of processors available. Moreover, the lower bound Ω(M/p+H+(H/p) log (H/p)) is presented for the parallel algorithms on distributed memory system, where M and H represent total number of the active nodes and that of the expanded nodes processed by p processors, respectively. In addition, a nearly fastest general parallel best-first B&B algorithm is put forward. The parallel algorithm is the fastest one as p = max{hε, r}, where ε = 1/ rootlogh, and r is the largest branch number of the nodes in the state-space tree.
文摘针对蒙特卡洛树搜索算法(Monte Carlo tree search,MCTS)收敛速度过慢,且在博弈过程中关键节点会出现信息丢失等问题,以中国象棋为载体,构建适用于中国象棋博弈系统的策略价值网络,提出了一种基于统计数据的并行蒙特卡洛树搜索算法(parallel Monte Carlo tree search based on statistics,SPMCTS)。将并行化的重点设置在MCTS四个步骤中最耗时的扩展和模拟步骤,有效避免了算法执行过程中的等待时差。并且引入一组新统计数据,这些数据用于在MCTS的选择步骤中修改节点的选择策略,保证在进行节点选择时获取和利用更多的可用信息,缓解信息丢失对精度造成的影响。实验结果表明,与现有并行蒙特卡洛树算法相比,SPMCTS在搜索速度上加快了约34%,且在对弈实验中,博弈胜率也能保持在80%左右。验证了SPMCTS的有效性。
文摘在静态栅格地图中,针对传统蚁群算法进行AGV(Automated Guided Vehicle,自动引导车)路径规划收敛慢且搜索结果容易陷入局部最优的问题,提出一种融合跳点搜索(Jump Point Search,JPS)和双向并行蚁群搜索的改进算法.首先,对实际研究环境进行栅格化建模,使用改进的跳点搜索算法生成双向搜索的初始次优路径,为双向蚁群搜索提供初始搜索方向参考.其次,在双向并行蚁群搜索过程中采用改进的转移概率启发函数,该函数在确定下一个转移节点时考虑了避免AGV与障碍物碰撞的因素,同时通过设计信息素共享机制并结合改进的信息素增量及浓度两种融合模型,共享和更新全局信息素浓度,以更好地探索和优化路径,保证双向路径连结.最后,与传统蚁群算法进行实验结果对比,验证了改进算法的全局搜索能力、效率和安全性.
