Statistics Theory and Methodsmathscidoc:1912.43298
2015.12
Children learn effortlessly by example and exhibit a remarkable capacity of generalization. The field of machine learning, on the other hand, stumbles along clumsily in search of algorithms and methods, but nothing available today comes even close to an average two-year-old toddler. So, modestly, we present some results on one of the main paradigms in machine learningnearest neighbor methods. Rummaging through old data for the closest match seems like a sensible thing to do, and that primitive idea can be formalized and made rigorous. In the field of nonparametric statistics, where one is concerned with the estimation of densities, distribution functions, regression functions, and functionals, the nearest neighbor family of methods was in the limelight from the very beginning and has achieved some level of maturity.
@inproceedings{jianqing2015lectures,
title={Lectures on the nearest neighbor method},
author={Jianqing Fan, and Luc Devroye},
url={http://archive.ymsc.tsinghua.edu.cn/pacm_paperurl/20191221113548303141858},
year={2015},
}
Jianqing Fan, and Luc Devroye. Lectures on the nearest neighbor method. 2015. http://archive.ymsc.tsinghua.edu.cn/pacm_paperurl/20191221113548303141858.