Importance Resampling for neural model selection

• Main Author: 張泰隆
• Other Authors: 洪文良
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/73998003671462917473

碩士 === 國立新竹教育大學 === 數學教育學系碩士班 === 93 === Bootstrap techniques, resampling computation techniques, have introduced new advances in model evaluation (Bootstrap for neural model selection, Riadh Kallel [1]). Using resampling methods to construct a series of new samples which are based on the original data set, allows to estimate the stability of the parameters. However two main disadvantages must be outlined. First, if or is high, computation time can be very long. Secendly, Efron (1987) finds that reasonable standard error estimates are obtained with only . In this paper, we use importance resampling method for neural model selection. These examples indicate that it is better than uniform resampling method. Therefore it decreases the number of replications and computation time.