求解多目标优化问题的自适应混沌混合蛙跳算法

针对多目标优化问题提出一种自适应混沌混合蛙跳算法MACSFLA(Adaptive chaos shuffled frog leaping algorithm for multiobjective optimization)。使用动态权重因子策略以提高混合蛙跳算法SFLA(Shuffled Frog Leaping Algorithm)收敛效率,引入基于Pareto支配能力的SFLA子族群划分策略,使得SFLA能够应用于多目标优化问题。在此基础上,MACSFLA首先利用SFLA快速寻优能力接近理论Pareto最优解,然后采用自适应网格密度机制动态维护外部存储器Pareto最优解规模,并使用自适应混沌优化技术改善Pareto最优解集样本多样性,最后利用Pareto最优解选择策略为青蛙种群选择最优更新粒子。多目标函数测试实验结果表明,与MOPSO和NSGA-Ⅱ相比,MACSFLA在Pareto最优解集均匀性和多样性上有明显优势。     

EV-DO监测系统中R-P接口业务信令追踪的研究与实现

在EV-DO网络中,信令监测系统对于用户数据业务运营的分析尤为重要。针对EV-DO网络中的重要数据接口R-P接口间A11及PPP信令CDR(Call Data Record)的合成,结合主叫和寻呼过程的业务特征,提出一系列优化及改进方案。经过EV-DO现网数据的测试以及产生的相应实验结果表明,优化信令与业务数据消息间的关联机制,缩短了数据网络寻呼与用户主叫业务信令的追踪统计时间最高可达30 s,指标的细化提升了故障定位准确性。这项改进方案对于优化网络运维的分析平台有着重要意义。     

一种基于蚁群优化的图像分类算法

现有图像降维方法中特征信息被过多压缩,从而影响图像分类效果。提出IC-ACO算法,利用蚁群算法来解决图像分类问题。算法充分提取并保留图像的各种形态特征。利用蚁群优化算法在特征集中自动挖掘有效特征和特征值,构建各类分类规则,从而实现图像的分类识别。在真实的车标图像数据集上的实验结果表明,IC-ACO算法比其他类似算法具有更高的分类识别率。     

小数据集条件下的多态系统贝叶斯网络参数学习

针对贝叶斯网络中多父节点条件概率分布参数学习问题,提出了一种适用于多态节点、模型不精确、样本信息不充分情形的参数学习方法.该方法利用因果机制独立假设,分解条件概率分布,使条件概率表的规模表现为父节点个数和状态数的线性形式;利用Leaky Noisy-MAX模型量化了多态系统模型未含因素对参数学习的影响;从小样本数据集中获取模型参数并合成条件概率表.结果表明,该方法能提高参数学习效率与精度.     

蝙蝠算法优化神经网络的网络入侵检测

研究网络入侵安全问题,网络入侵具有隐蔽性、随机性和突发性等复杂变化特点,传统方法无法描述变化规律,导致入侵检测正确率低。为了提高网络入侵检测效果,针对BP神经网络参数优化问题,提出一种蝙蝠算法优化BP神经网络的权网络入侵检测模型(BA-BPNN)。首先将BP神经网络参数编码为蝙蝠个体,并以网络入侵检测正确率作为个体适应度函数,然后通过模拟蝙蝠飞行过程找到BP神经网络最优参数,最后根据最优参数建立网络入侵检测模型。在Matlab 2012平台采用KDD CUP 99数据集仿真测试,结果表明,BA-BPNN解决了传统神经网络模型存在的难题,提高网络入侵检测正确率。     

Chaotic bat swarm optimisation (CBSO)

Bat swarm optimisation (BSO) is a novel heuristic optimisation algorithm that is being used for solving different global optimisation problems. The paramount problem in BSO is that it severely suffers from premature convergence problem, that is, BSO is easily trapped in local optima. In this paper, chaotic-based strategies are incorporated into BSO to mitigate this problem. Ergodicity and non-repetitious nature of chaotic functions can diversify the bats and mitigate premature convergence problem. Eleven different chaotic map functions along with various chaotic BSO strategies are investigated experimentally and the best one is chosen as the suitable chaotic strategy for BSO. The results of applying the proposed chaotic BSO to different benchmark functions vividly show that premature convergence problem has been mitigated efficiently. Actually, chaotic-based BSO significantly outperforms conventional BSO, cuckoo search optimisation (CSO), big bang-big crunch algorithm (BBBC), gravitational search algorithm (GSA) and genetic algorithm (GA).     

Variable Structure Control with chattering elimination and guaranteed stability for a generalized T-S model

In this paper, a fuzzy based Variable Structure Control (VSC) with guaranteed stability is presented. The main objective is to obtain an improved performance of highly non-linear unstable systems. The main contribution of this work is that, firstly, new functions for chattering reduction and error convergence without sacrificing invariant properties are proposed, which is considered the main drawback of the VSC control. Secondly, the global stability of the controlled system is guaranteed.The well known weighting parameters approach, is used in this paper to optimize local and global approximation and modeling capability of T-S fuzzy model.A one link robot is chosen as a nonlinear unstable system to evaluate the robustness, effectiveness and remarkable performance of optimization approach and the high accuracy obtained in approximating nonlinear systems in comparison with the original T-S model. Simulation results indicate the potential and generality of the algorithm. The application of the proposed FLC-VSC shows that both alleviation of chattering and robust performance are achieved with the proposed FLC-VSC controller. The effectiveness of the proposed controller is proven infront of disturbances and noise effects.     

基于MQP模型的电离图参数反演解析算法研究

针对现今普遍采用的偏导迭代电离图反演算法存在的求解依赖于初值选择的不足,给出了一种初值选择准则,保证了该算法的收敛性。同时,利用该准则,以及在对电离图虚高表达式进行合理近似的基础上,提出了一种新的电离图参数反演算法,该方法实现电离图参数反演的解析计算,较偏导迭代法计算量大大减少,更好满足工程实时性要求。     

一种基于改进支持向量机的文本倾向性分类算法

文本的倾向性分类器是文本倾向性分类的核心部分,它用于将待分类的文本映射到某一倾向性类别中去。传统支持向量机的核函数学习能力和泛化推广能力的平衡性有待提高,而且参数选择不易。对目前文本倾向性分类算法使用的传统的支持向量机进行了改进,一是构造了多核函数;二是使用粒子群算法对支持向量机的参数进行优化,平衡了核函数的全局性和局部性,更有利于对样本数据的学习和推广;最后利用改进的支持向量机构造文本倾向性分类算法。