Research of classification methods of hyper-spectral data, elucidation of the theoretical nature and applications to real data

近年来，需要一种适用于超维数据的分类方法。我们提出了一种袋式AdaBoost方法，它可以给出复杂的决策边界，避免了过度学习。通过马尔可夫随机场的空间相关性建模，研究了与地理空间数据相关的分类方法和回归问题，以及土地覆盖类别的分解。此外，我们还出版了一本书，讨论了统计模型选择的信息标准。

Recently, classification methods applicable to hyper-dimensional data are required. We proposed a bagging-type AdaBoost method, which can give complicated decision boundaries and avoid over learning. Classification methods and regression problems related to geo-spatial data, and unmixing of land-cover categories were also studied through modeling of spatial dependency by Markov random fields. Furthermore, we published a book discussing information criteria for statistical model selection.