改进的核极限学习机定位算法

针对神经网络无线定位方法，存在训练耗时长，定位结果易受噪声干扰的问题，提出了一种改进的核极限学习机无线定位算法。采取在同一位置进行多次测量的方法得到训练数据；把同一位置测得的数据划分为一个样本子空间并提取样本子空间的特征，以样本子空间的特征代替原来的训练数据；利用矩阵近似及矩阵扩展的相关理论改进核极限学习机算法；将处理过的训练数据利用改进的核极限学习机进行训练，得到定位预测模型。仿真结果表明，在相同数据集下，改进的核极限学习机训练用时短、定位速度快；在相同噪声干扰情况下，此算法定位预测误差小。经验证，该算法不但能提高网络的训练速度、定位速度，还能有效地降低噪声的干扰，提高定位精度。

Improved kernel extreme learning machine localization algorithm

Aiming at the problems of neural networks wireless location, such as large training time consumption and positioning results easily interfered by noise, this paper presents an improved kernel extreme learning machine wireless positioning algorithm. Firstly, the training data is obtained by the method of multiple measurements at the same location. Then, the data obtained at the same position is divided into a sample subspace and the characteristics of the sample subspace are extracted to replace the original training data. At the same time, the kernel extreme learning machine algorithm is improved by using the matrix approximation and matrix extension theory. Finally, the processed training data is trained by the improved kernel extreme learning machine, and the positioning prediction model is obtained. The simulation results show that the improved kernel extreme learning machine has shorter training time and the positioning speed is faster under the same data set. In the case of the same noise interference, the algorithm makes less prediction errors. It has been proved that the algorithm can not only improve the training speed and positioning speed of the network, but also reduce the interference of noise and improve the positioning accuracy.