贝叶斯证据框架下的支持向量机及其在陀螺漂移预测中的应用

The Application of Support Vector Machines Within Bayesian Evidence Framework to Drift Prediction of Gyro

将贝叶斯证据框架引入到标准支持向量机中去,研究了贝叶斯证据框架下标准支持向量机的参数调整方法,并具体研究了高斯核情况下参数的调整.利用获得的结论对陀螺漂移数据进行了预测.结果表明,该方法为参数的选取提供了一个统一的框架,克服了在样本改变的情况下标准回归支撑矢量机方法中某些参数只能凭经验通过交叉验证的方法进行频繁的手动调整的不足,且具有较高的精度.

Bayesian evidence framework is introduced into the standard support vector machines （SVM）.A selection and tuning method of parameter within Bayesian evidence framework is proposed for standard SVM with Gauss kernel. The method is used to predict the drift of gyro. The result shows that a unique framework is provided for the parameter selection by the method. The drawback of frequent manual parameter tuning of standard SVM is overcome through cross validation with experience when the sample is changed , and the accuracy of the method is higher.