论文部分内容阅读
特征选择是模式识别的关键步骤之一。本文采用“留一交替”的无偏考试方法，充分利用过并的已知样本，从大量的地震初始特征中选择出错误率最小的特征组合，并进一步对选择出的特征的物理意义进行分析，以指导地震解释工作。该方法在Sun工作站上实现，取得了良好效果。
Feature selection is one of the key steps in pattern recognition. In this paper, we use the unbiased test-taking method of “leaving one after another alternative” to make full use of the known samples and select the feature combination with the lowest error rate from the large number of initial earthquakes and further analyze the physical meaning of the selected features , To guide the earthquake interpretation work. The method is implemented on a Sun workstation and achieved good results.