为了缓解Internet网络拥挤状况,提高用户访问网站的响应速度,从技术上解决由于网络带宽小、用户访问量大、网点分布不均等原因所造成的用户访问网站响应速度慢的问题,提出了一种新的缓存资源分配方法——细胞优化算法。该算法是模仿自然细胞系统功能的一种智能优化方法,其通过模拟细胞内部结构和原理,对细胞核、细胞质的浓度、细胞间的亲和度、细胞优化机制、细胞的动态演化过程建立数学模型。给出了算法的并行计算结构和步骤。最后,通过理论证明、仿真实验与同类算法的比较,验证了算法求解CDN缓,存资源分配问题的有效性。

计算机安全系统与生物免疫系统具有很多的相似性,它们都需要在不断变化的环境中维持自身的稳定性。提出复合免疫算法,并应用到入侵检测系统中,以保护网络安全。针对经典的人工免疫算法在性能上存在的缺陷进行了改进,完善了其核心算法——否定选择算法,在否定选择算法中加入了分段技术和关键位,避免了恒定的匹配概率导致的匹配漏洞,降低了系统漏检率。并将遗传算法中的克隆选择算法和改进的否定选择算法结合为复合免疫算法,提高了检测器生成的动态性和多样性。最后,通过数学理论分析与仿真实验模拟,验证了改进算法的有效性和可行性,并且与其它经典算法进行了比较,结果证明,改进算法可以提高系统性能。

五行学说蕴含信息动力学,然而在网络中却一直未被很好地利用,为此提出一种基于五行原理的五行粒子模型方法来求解多Agent系统的分布式问题。五行粒子模型可以很好地描述和处理多Agent系统中Agent之间存在的随机、并发、多类型的交互行为。基于五行粒子模型内部存在的生克关系,以及五行自身蕴含的稳定性和平衡性,对五行粒子模型和多Agent系统分布式问题求解进行探讨,并对多Agent系统中的各个Agent进行行为建模,进而提出多Agent系统分布式问题求解的五行粒子模型算法。最后,通过实验验证了该算法的有效性。

撒谎行为的存在会破坏CDN缓存分配的公平性。 使用博弈论对服务器在缓存分配过程中的自私撒谎行为进行了研究。经分析发现,服务器撒谎行为的本质就是当缓存不足时,额外多申请一定量缓存；而当缓存充足时,则诚实地申请所需缓存量。针对这种撒谎行为,提出了一种公平分配算法,在计算服务器的缓存申请量时,考虑其历史缓存申请量,并根据不同阶段申请量的有效性不同引入年龄因子,(重新)计算得到服务器的当前有效缓存申请量,使得撒谎的服务器与诚实的服务器相比受到更多损失,以此来促使其停止撒谎行为。同时,公平算法还保证了系统的最大吞吐量,并引入了价格机制来保证诚实的服务器得到更高的需求满足度。仿真实验结果表明,公平算法对于上述撒谎行为有很好的改善效果。

A wolf pack can hunt prey efficiently due to reasonable social team hierarchy and effective team cooperation. Inspired by the collective intelligence of wolf pack, in this paper, a novel swarm algorithm named the social team building optimization (STBO) algorithm is proposed for solving optimization problems. In order to mimic the method of social team building, which is an optimization process in reality, STBO algorithm is in terms of social team hierarchy, team building state and process control. Firstly, the social team model separates individuals of population into different swarms according to the appropriate team hierarchy. In this way, the proposed algorithm not only has fast search speed but also avoids to fall into the local optimum prematurely. Secondly, the team building state model divides the optimization process into three states. In different states, individuals at different levels act diverse social behaviors to make the algorithm maintain population diversity and possess better search capability. Thirdly, the team power model is designed to determine the states of optimization process by means of the team power and the team cohesion. The main aim of this model is to make the algorithm have a good balance between exploration and exploitation, namely to find the optimal solutions as possible as it can. Moreover, the mathematical models of STBO are educed by the swarm theory, the state evolution theory and the energy–entropy theory. Meanwhile, the convergence property of the presented algorithm has been analyzed theoretically in this paper. And STBO was compared to three classical nature-inspired algorithms on 11 basic standard benchmark functions and also three state-of-the-art evolutionary methods on CEC2016 competition on learning-based single-objective optimization. Some simulation results have shown the effectiveness and high performance of the proposed approach.