欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
为了更好地逼近真实物理场景，对传统的多指数模型作了一些改进，将权因子设置为噪声方差平方的倒数，提出一种基于循环矩阵（CM）的算法用于估计衰减项数.为了求解上述改进模型，提出一种混沌免疫粒子群优化（CIPSO）算法.该算法将人工免疫系统中的克隆、交叉、变异和接收器修正算法嵌入粒子群算法中，并采用混沌算子实现变异，然后将惯性因子改为自适应变化.实验表明：提出的权因子设置更符合实际；用于估计项数的CM算法在估计精度与运行时间上均优于传统的ILS算法；CIPSO算法在收敛精度与运行时间上也优于传统的优化算法，如可信域法、LM法、高斯-牛顿法、差分进化算法和粒子群算法等.
英文摘要：
In order to approach practical physical scene, traditional multi-exponential model is improved as follows： the weight factor is set as the reciprocal of the square of noise variance, and a method based on circular matrix （CM） is proposed to estimate the number of decay terms. To efficiently solve this model, a novel algorithm called chaotic immune particle swarm optimization （CIPSO） is proposed. The operators of clone, crossover, mutation, and receptor editing are embedded into particle swarm optimization; the chaotic operator is used to realize mutation, and the inertial factor is stipulated as adaptive variation. Experiments demonstrate that the proposed setting of weight factor is more practical; the CM method is superior to traditional iterative least square （ILS） method in terms of estimation accuracy and computation time; the CIPSO algorithm outperforms traditional optimization methods such as the trust-region method, Levenberg-Marquardt （LM） method, Gaussian-Newton method, difference evolution algorithm, PSO algorithm in terms of convergence accuracy and computation time.
同期刊论文项目
同项目期刊论文
位置：