共查询到18条相似文献,搜索用时 109 毫秒
1.
2.
3.
差分进化算法是一种有效求解全局优化问题的方法,为进一步提高求解精度,加快求解过程,文中提出一种梯度策略自适应差分进化算法。该算法是在差分进化算法中加入梯度下降法,使其不仅有较好的全局搜索能力,且具有传统优化方法的快速局部搜索能力,因此具有较高搜索精度和较快的搜索过程。通过对CEC2005测试集中的1~14号测试函数进行仿真实验,并与SaDE,NSDE以及CMAES等算法实验结果进行了对比,结果表明了该算法的有效性。 相似文献
4.
5.
差分进化算法是一种结构简单、易用且鲁棒性强的全局搜索启发式优化算法,它可以结合约束处理技术来解决约束优化问题.机器学习在进化算法中,经常可以引导种群的进化,而且被广泛地应用于无约束的差分进化算法中,但对于约束差分进化算法却很少有应用.针对这一情况,提出了一种基于反向学习的约束差分进化算法框架.该算法框架采用基于反向学习的机器学习方法,提高约束差分进化算法的多样性和加速全局收敛速度.最后把该算法框架植入了两个著名的约束差分进化算法:(μ+λ)-CDE和ECHT,并采用CEC 2010的18个Benchmark函数进行了实验评估,实验结果表明:与(μ+λ)-CDE和ECHT相比,植入后的算法具有更强的全局搜索能力、更快的收敛速度和更高的收敛精度. 相似文献
6.
7.
差分进化算法(DE)已被广泛应用于解决稀疏面阵优化问题,针对DE 算法早熟、全局搜索能力差、容易陷于局部最优的问题,提出一种混合变异差分进化算法,通过加入概率因子来平衡算法收敛速度与全局搜索能力,以阵列孔径、阵元数量以及阵元间距为约束条件,将算法中的实数编码转化为二进制编码,以方向图平面峰值旁瓣电平之和最低为目标函数,通过优化后得到的阵元分布,得到稀疏优化阵列的三维方向图。仿真结果表明:该方法在满足约束条件的同时,能够避免算法早熟得到较优的目标函数值,概率因子为算法提供了额外的自由度。 相似文献
8.
阵列天线方向图综合作为智能天线的一项重要技术对于复杂系统的优化问题提出了更高的要求.本文结合差分进化的基本思想和遗传算法的基因交叉和变异方法构建了一种新的混合优化方法.该算法使用差分进化算子作为主要优化手段,结合遗传算法优胜劣汰的基本思想以及其中的交叉和变异方法,通过融合两种优化方法各自的优点,可以显著改善多参数、高度非线性问题的优化结果、提高计算效率.计算机仿真结果表明,与现有算法相比,该算法用于天线方向图综合具有收敛速度快、优化能力强、算法可靠等优势. 相似文献
9.
针对标准的差分进化算法只能处理连续空间的优化问题,提出了一种基于取整策略的差分进化算法。该方法只需要对优化变量进行四舍五入取整,就能够把标准差分进化算法用于稀疏阵列天线方向图优化。将取整策略的差分进化算法应用到六边形平面稀疏天线阵的布阵设计。为了计算六边形阵列天线的方向图,提出在口径中添加虚拟单元的计算模型,把六边形阵列转化为可以实现二维快速傅里叶变换的矩形阵列。以改善阵列峰值副瓣电平为目的进行仿真试验,结果表明,优化后的稀疏天线阵峰值旁瓣电平与采用遗传算法相比改善了4.5~5.1 dB,且具有计算速度快、稳定性好的优点。 相似文献
10.
11.
针对常规最大类间方差法在多阈值图像分割中存在的运算量大、计算时间长、分割精度较低等问题,该文提出一种基于改进的自适应差分演化(JADE)算法的2维Otsu多阈值分割法。首先,为增强初始化种群的质量、提升控制参数的适应性,将混沌映射机制融入到JADE算法中;进而,通过该改进算法求解2维 Otsu 多阈值图像的最佳分割阈值;最终,将该算法与差分进化(DE), JADE,改进正弦参数自适应的差分进化(LSHADE-cnEpSin)以及增强的适应性微分变换差分进化(EFADE) 4种算法的2维Otsu多阈值图像分割进行比较。实验结果表明,与其它4种算法相比,基于改进JADE算法的2维Otsu多阈值图像分割在分割速度以及精度上均有较明显的改善。 相似文献
12.
Reconfigurable antenna arrays are often capable of radiating multiple patterns by altering the excitation phases of the array elements. In this paper, an efficient method based on FFT is presented for generating dual-radiation pattern from a single rectangular planar array by modifying the excitation phases of the array elements while sharing common amplitudes. The common amplitudes shared by both the patterns and the phases which play the role of turning between the two patterns when updated over zero phase among the elements are computed using Self-adaptive Differential Evolution (SaDE) algorithm. Two different beam-pairs of pencil/pencil and pencil/flat-top are generated from the proposed array while maintaining precise design specifications. The proposed method greatly reduces the computational time as compared with the conventional method for calculating beam patterns. The dynamic range ratio of the excitation amplitudes are kept below a threshold level to reduce the design complexity of the attenuators at the feed network level and to minimize the effect of mutual coupling among the array elements. To illustrate the effectiveness of SaDE, the fitness functions associated with the two beam-pairs are minimized individually using differential evolution (DE) algorithm and particle swarm optimization (PSO) algorithm. Results clearly show the superiority of SaDE over DE and PSO to handle the proposed problem. 相似文献
13.
A novel Cauchy mutated cat swarm optimization (CMCSO) that features effective global search capabilities with fast convergence is introduced in this paper. The Cauchy mutation enables the cats of the cat swarm optimization (CSO) algorithm to seek their positions in directions that avoid the problem of premature convergence and local optima. In this communication, CMCSO is applied to the synthesis of linear aperiodic arrays for minimizing sidelobe level and controlling the null positions. Various synthesis examples are considered and the obtained results are compared with linear aperiodic array designs from literature. Numerical results demonstrate that the proposed method is superior to existing methods in terms of accuracy and convergence speed. Some of the synthesized aperiodic array designs are implemented with wire dipole antenna elements using a full-wave electromagnetic simulator. Furthermore, experiments are conducted on several standard benchmark complex multimodal problems to demonstrate the effectiveness of the proposed method. The sensitivity analysis is performed on different parameters of CMCSO to demonstrate their influence on the overall performance of the benchmark and antenna array synthesis problems. 相似文献
14.
15.
The advantages of using diversely polarized antennas to determine bearings of multiple cochannel narrow-band signals are shown. Three bearing estimation algorithms-maximum likelihood (ML), adapted angular response (AAR), and Music (multiple signal classification)-are extended to handle antenna arrays with diverse polarizations; the maximum entropy method does not readily extend. The proposed algorithms are applicable to arbitrary antenna locations and directional characteristics and arbitrary noise correlations between the antenna outputs. The algorithms are compared on the basis of multiple signal resolution and bearing accuracy in the presence of noise. The Music algorithm exhibits superior performance at moderate to low signal-to-noise ratio (SNR). 相似文献
16.
17.
18.
Selecting array configurations for MIMO systems: an evolutionary computation approach 总被引:2,自引:0,他引:2
Karamalis P.D. Skentos N.D. Kanatas A.G. 《Wireless Communications, IEEE Transactions on》2004,3(6):1994-1998
This paper presents an antenna selection method for multiple-input multiple-output wireless systems. By exploitation of the channel transfer matrix, the antenna selection criterion is the maximization of the instantaneous capacity achieved using a specific number of transmitting and receiving antenna array elements. For each environment, the proposed method applies a genetic algorithm which seeks the most advantageous subset of antenna elements. The results are based on measured and simulated channels and show that the proposed method selects array configurations that yield superior performance compared to the arrays usually employed. Furthermore, comparative analysis results are presented, with respect to a state-of-the-art algorithm. 相似文献