关于ε不敏感损失函数推广误差的界

GENERALIZATION ERROR BOUND ABOUT ε INSENSITIVE LOSS FUNCTIONS

本文研究了学习理论中推广误差的界的问题.利用ε不敏感损失函数的性质,分别获得了逼近误差和估计(样本)误差的界,并在特定的假设空间上得到了学习算法推广误差的界.

In the article,we investigate the bound of the generalization error in learning theory.By using the property ofεinsensitive loss function,we establish the estimates of the approximation error and the estimation error respectively,then in specific hypothesis space,we get the bound of the generalization error of learning algorithm.