The Generalization Performance of ERM Algorithm with Strongly Mixing Observations
首发时间:2007-11-27
Abstract:The generalization performance is the main purpose of machine learning theoretical research. The previous main bounds describing the generalization ability of ERM algorithm are based on independent and identically distributed (i.i.d.) samples. When the complexity of the given function set is high, the problem of solving the ERM algorithm is usually ill-posed and overfitting may happen. In order to study the generalization performance of ERM algorithm with dependent observations, in this paper we decompose firstly the given function set into different compact subsets, and then we establish the exponential bound on the rate of relative uniform convergence on these compact subsets for ERM algorithm with strongly mixing observations. In the end, we obtain the bounds on the generalization ability of ERM algorithm with strongly mixing observations over the given function set, which extend the previous results of i.i.d. observations to the case of strongly mixing observations.
点击查看论文中文信息
混合样本ERM算法的推广性能
摘要:The generalization performance is the main purpose of machine learning theoretical research. The previous main bounds describing the generalization ability of ERM algorithm are based on independent and identically distributed (i.i.d.) samples. When the complexity of the given function set is high, the problem of solving the ERM algorithm is usually ill-posed and overfitting may happen. In order to study the generalization performance of ERM algorithm with dependent observations, in this paper we decompose firstly the given function set into different compact subsets, and then we establish the exponential bound on the rate of relative uniform convergence on these compact subsets for ERM algorithm with strongly mixing observations. In the end, we obtain the bounds on the generalization ability of ERM algorithm with strongly mixing observations over the given function set, which extend the previous results of i.i.d. observations to the case of strongly mixing observations.
关键词: 混合样本, ERM算法, 推广误差
论文图表:
引用
No.1659211092911961****
同行评议
勘误表
混合样本ERM算法的推广性能
评论
全部评论0/1000