一般化状態空間モデルと自己組織化の方法(<論文特集>「情報論的学習理論(IBIS2000)」)

抄録

For automatic extraction of essential information and discovery from massive time series, it is necessary to develop a method which is flexible enough to handle actual phenomena in real world.That can be achieved by the use of general state space model, and it provides us with a unified tool for analyzing complex time series.To apply these general state space models, development of practical filtering and smoothing algorithms is indispensable.In this article, the non-Gaussian filter/smooother, Monte Carlo filter/smoother and self-organizing state space model are shown.As applications of the method, problems of detecting sudden changes of the trend and nonlinear smoothing are shown.