基于最小二乘支持向量机的阵列波束优化研究

针对标准支持向量机计算复杂度高、内存开销大、训练速度慢的缺点,为改善标准支持向量机的训练效率,快速优化阵列波束,提出了基于最小二乘支持向量机(least squares support vector machine,LSSVM)的阵列波束优化方法。LSSVM采用二次损失函数取代标准支持向量机中的不敏感损失函数,将不等式约束变为等式约束,从而将二次规划问题转化为一个线性矩阵求解问题,具有良好的快速性;与传统的标准支持向量机波束形成相比,所需计算资源更少,训练速度更快,计算效率更高,泛化能力更强。仿真实验结果表明:在保持波束形成的性能指标基本不变的情况下,LSSVM降低了计算复杂度,减少了内存开销,提高了运算速度和收敛精度,为波束形成器的优化设计提供了一种新的有效方法。

Research on Optimization of Array Beamforming based on Least Squares Support Vector Machine

For the shortcomings of standard support vector machine, such as high computational complexity, memory overhead of large and slow training, in order to improve the training efficiency of standard support vector machine and to optimize the beamforming rapidly it proposes the optimization method of array beamforming based on Least Squares Support Vector Machine (LSSVM). The method based on LSSVM utilizes quadratic loss function to replace the insensitive loss function of support vector machine and changes the inequality constraints into equality constraints. Thus it transforms a quadratic programming problem into a solving linear matrix problem and has a nicer speediness. Compared with the traditional standard support vector machine beamforming, it requires less computational resources and has stronger generalization ability and higher computational efficiency with faster training. The simulation results show that the method of LSSVM reduces the computational complexity and the memory occupancy, increases computing speed and convergence accuracy and it provides a new and effective ways for the optimization design of beamformer.