Optimization of Regression Models based on the Analysis of Risk Functions

R.M. Kil and I. Koo (Korea)


regression, incremental learning, VC dimension, general ization, optimization


This paper presents a new method of optimizing the struc ture of regression models based on the analysis of risk func tions. The goal of regression models is minimizing the true risk (or general error) for the whole distribution of sample space. However, the true risk can not be estimated from the finite number of samples which are usually given for the learning of regression models. In this sense, we investi gate the functional forms of true risk bounds and suggest an estimation method of the performance of regression mod els. The optimal structures of regression models can be determined from the estimated performance of regression models.

Important Links:

Go Back