U.S.-Brazil Cooperative Research: Multivariate Bayesian Classification Methods in Spatial Statistics

该奖项将支持加州大学河滨分校的James Press教授和巴西巴西国家科学计算实验室（LNCC）的Ruben Klein教授的合作研究。本研究的目的是研究空间统计中的多元贝叶斯分类方法。使用迭代贝叶斯和经验贝叶斯程序的遥感数据的图像重建和上下文分类将为这次合作增加一个国际层面。研究人员已经就这一主题共同撰写了许多期刊文章，他们计划将现有模型扩展到更广泛的背景下，并以各种方式改进当前模型。将对模拟数据和实际数据进行研究。他们建议使用自然共轭先验族（同时使用迭代过程进行预测分类）研究“块金效应”模型、各种空间相关结构和自适应起始值模型。结果应适用于农业和地球物理背景下的卫星图像，图像重建和模式识别问题。

This award will support collaborative research by Professor James Press of the University of California at Riverside and Professor Ruben Klein of the National Laboratory for Scientific Computation (LNCC) in Rio de Janeiro, Brazil. The aim of the research is to study multivariate Bayesian classification methods in spatial statistics. Image reconstruction and the contextual classification of remotely sensed data using iterative Bayesian and empirical Bayesian procedures will add an international dimension to this collaboration. The researchers, who have coauthored a number of journal articles on this topic, plan to extend existing models to broader contexts and to improve current models in various ways. Both simulated and actual data will be studied. They propose to study "nugget effect" models, various spatial correlation structures, and adaptive starting value models using natural conjugate prior families (while using iterative procedures for predictive classification). Results should be applicable to satellite imagery in agricultural and geophysical contexts, image reconstruction, and problems in pattern recognition.