Mathematical Methods to Extract Information from High-dimensional Data (E01)

为了开发一种识别高维时间序列中事件的通用数学方法，工作程序结合了三种不同的近似方法。这些都是基于一个分析的数值方法的函数逼近和降维，统计聚类的随机方法和基于离散图的方法生成一个有限的状态转换图模型。将这些技术结合在一起，目标是提出新的算法方法，回答理论和实践问题，并支持人类可解释的时间序列分析。

With the aim to develop a general mathematical methodology for identifying events in high dimensional time series, the work program combines three different approximation methods. These are based on an analytic numerical approach for function approximation and dimension reduction, a stochastic approach for statistical clustering and a discrete graph-based method for generating a finite state transition graph model. In bringing these techniques together, the goal is to come up with new algorithmic methods that answer both theoretical and practical questions and support human-interpretable time series analysis.