论文部分内容阅读
针对实际应用中数据的批量到达,以及系统的存储压力和学习效率低等问题,提出一种基于信念修正思想的SVR增量学习算法.首先从历史样本信息中提取信念集,根据信念集和新增数据的特点选择相应的信念集建立支持向量回归模型并进行预测;然后对信念集进行修正,调整当前认知状态,使该算法对在线和批处理增量学习都有很好的适应性.在标准数据集上的测试验证了算法的良好性能;在某机场噪声实测数据上的对比实验也表明,该算法的性能明显优于传统学习算法和一般增量学习算法.
In order to solve the problem of batch arrival of data in practical application, as well as the storage pressure and low learning efficiency of the system, an SVR incremental learning algorithm based on belief revision is proposed.First, belief sets are extracted from historical sample information, Then, the belief set is modified and the current cognitive state is adjusted, so that the algorithm has good adaptability to online and batch incremental learning. The test on the standard dataset verifies the good performance of the algorithm. The comparative experiment on the measured data of an airport noise also shows that the performance of the algorithm is obviously better than the traditional learning algorithm and the general incremental learning algorithm.