粗糙随机样本的结构风险最小化原则

The structural risk minimization principle of rough-stochastic samples

介绍了粗糙随机理论的基本内容,提出了退火熵、生长函数和VC维等概念.并在此基础上构建了基于VC维的学习过程一致收敛速度的界;随后给出了粗糙随机样本的结构风险最小化原则;最后证明了在此原则下收敛速度渐进分析的一些相关性质.

First,the basic content of rough stochastic theory is introduced,and the concepts of annealed entropy,growth function and VC dimension are proposed.On the basis of these,the bounds on the rate of uniform convergence of learning process based on VC dimension are constructed.Then the idea of the structural risk minimization principle of rough-stochastic sample has been presented.Finally,some properties of asymptotic analysis about convergent rate based on this principle have been proved.