【摘要】 核机器学习算法是近几年发展起来的一类新的非线性技术,核Fisher判别分析是其中之一。核Fisher判别分析是经典Fisher线性判别基于核函数的非线性推广,并在实际资料的分类中取得明显效果。本文简化了核Fisher判别分析的计算过程,并将其用于油气储集层横向预测。两个实际资料的计算结果表明,在油气储集层横向预测中,核Fisher判别技术的性能优于Fisher线性判别、模糊模式识别和反向传播人工神经网络。更多还原

【Abstract】 The kernel-based machine learning algorithms proposed in last few years are a novel class of non-linear techniques, one of which is kernel Fisher discriminanting analysis. The kernel Fisher discriminant analyses is the non-linear generalization of classic Fisher linear discriminant based on the kernel functions and obtains the better results in the classification problem of real data. The paper introduced a lateral prediction approach of hydrocarbon reservoir based on kernel Fisher discriminanting analysis. The computational results of field data show that the performance of kernel Fisher discriminant is superior to those of Fisher linear discriminant, fuzzy pattern recognition and backward propagation neural networks in lateral prediction of hydrocarbon reservoir.更多还原