快速SVM算法及在陀螺仪参数漂移预测中的应用

针对支持向量机在对海量训练样本进行训练时,训练速度慢而导致难以应用的问题,通过分析训练样本数目与训练时间之间的关系,利用支持向量机对小样本学习的良好特性,提出了基于样本分组的支持向量机快速训练算法.将海量样本分成小样本进行训练,然后对训练得到的多个支持向量机进行加权处理得到决策函数.此方法在标准数据以及陀螺仪参数漂移数据上进行了仿真应用,方针结果证明该方法可大幅提高训练速度,同时保证了较好的泛化能力.

Fast Algorithm of SVM and Application to Parameter Drift Forecasting of Gyro

It is very slowly to train large number of training data points.For this problem,it analyzes the relation between the number of training data points and the time of training,and the divided training method was presented.This method divide the training samples into some subclasses which have less number of training samples,and then construct the classifying function with the weighted averages method.This method was tested on three benchmark data and parameters drift data of gyroscope.According to the simulat...