Partially-Observed Discrete Dynamical Systems
Partially-Observed Discrete Dynamical Systems
复制标题
DOI：
10.23919/acc50511.2021.9483049
复制
发表时间：
2021-05
期刊：
影响因子：
--
通讯作者：
Mahdi Imani;Seyede Fatemeh Ghoreishi
中科院分区：
文献类型：
--
作者：
Mahdi Imani;Seyede Fatemeh Ghoreishi
介绍了一种新的信号模型--部分观测离散动力系统(PODDS)。该信号模型是隐马尔可夫模型(HMM)的特例，其中状态是包含系统不同组件的信息的矢量，并且每个组件从有限实值集合中获取其值。该信号模型目前被视为有限状态隐马尔可夫模型，其中最大后验概率(MAP)准则用于状态估计。利用PODDS中状态变量的离散结构，提出了最优分量MAP(CMAP)状态估值器，给出了每个状态变量的MAP解。给出了计算最优估计器的完全递归过程，然后介绍了PODDS模型的一个具体实例，该模型适用于通过有噪声的时间序列数据观察到的监管网络。通过对随机监管网络的PODDS模型的数值实验，证明了该估计器的高性能。
This paper introduces a new signal model called partially-observed discrete dynamical systems (PODDS). This signal model is a special case of the hidden Markov model (HMM), where the state is a vector containing the information of different components of the system, and each component takes its value from a finite real-valued set. This signal model is currently treated as a finite-state HMM, where maximum a posteriori (MAP) criterion is used for state estimator purpose. This paper takes advantage of the discrete structure of the state variables in PODDS and develops the optimal componentwise MAP (CMAP) state estimator, which yields the MAP solution in each state variable. A fully-recursive process is provided for computation of this optimal estimator, followed by introducing a specific instance of the PODDS model suitable for regulatory networks observed through noisy time series data. The high performance of the proposed estimator is demonstrated by numerical experiments with a PODDS model of random regulatory networks.