针对拉普拉斯分类器的核参数选择问题,通过首先假设窗的三个估计中核参数取不同的值,然后运用智能遗传算法对核参数进行优选,得到一种新的分类器——智能拉普拉斯分类器。多个基准数据集上的实验结果证明,智能拉普拉斯分类器相对普通拉普拉斯分类器和支持向量机而言,具有较高的分类精度和稳定性,是一种有效的分类方法。
Supposing the kernel parameters of the three Parzen window-based estimators are different and using intelligent genetic algorithm (IGA) to optimize these parameters, a novel classifier is developed the intelligent Laplacian classifier (TILC). The experimental results in several benchmark datasets indicate that the proposed TILC achieve higher classification accuracy and stable than the Laplacian classifier and standard support vector machine (SVM). Consequently, TILC provides a promising ahernative for classification.