一种基于互信息的特征跃迁示例学习法

给出一种能够接受特征及变化的示例学习方法。该方法是对ID3方法的一种改进，传统ID3方法是基于持征值的学习，训练示例是若干组静态特征值，其局限性在于不能理解和记忆特征的变化信息，尤其没考虑特征间的动态相关。改进后的方能学习动态特征，接受的训练示例是特征值在一定间隔内的初值和终值，从中获取特征值及其在指定间隔的跃迁，该方法能够学习数据动态趋势，尤其能够挖掘出特征间动态相关。通过若干例子测试，该方法适用于具有多元动态相关特征问题的分类。

A METHOD OF LEARNING FROM EIGENVALUE TRANSITION OF EXAMPLES BASED ON INFORMATION GAIN

A method of learning eigenvalues and their varieties is given. It improves the method ID3 which learns from static eigenvalues of examples. The varieties and dynamic correlation of eigenvalues are ignored in the method ID3. In the given method, the change of data can be learned because the training data is the initial and end eigenvalue of the interval. All eigenvalue's varieties and correlation can be understood and remembered. The method can be applied as classifier when the multi-parameters are dynamically associated.