Voronoi tree models for distribution-preserving sampling and generation
作者：
Highlights：
• A method to learn a nonparametric model of a dataset through Voronoi binary trees.
• Models built according to a distance measure to preserve the original distribution.
• The Voronoi models can be used either for sampling or generation of new data.
• Applicable to different tasks, like batch selection, imbalanced sets, reconstruction.
• Simulation tests show the good performance of the method in various learning tasks.
摘要
•A method to learn a nonparametric model of a dataset through Voronoi binary trees.•Models built according to a distance measure to preserve the original distribution.•The Voronoi models can be used either for sampling or generation of new data.•Applicable to different tasks, like batch selection, imbalanced sets, reconstruction.•Simulation tests show the good performance of the method in various learning tasks.
论文关键词：Voronoi tree models,Sampling,Generative models,Density estimation,Noparametric models
论文评审过程：Received 6 March 2018, Revised 22 May 2019, Accepted 13 August 2019, Available online 14 August 2019, Version of Record 23 August 2019.
论文官网地址：https://doi.org/10.1016/j.patcog.2019.107002
