An optimization based state and parameter estimation method is presented where the required Jacobian matrix of the cost function is computed via automatic differentiation. Automatic differentiation evaluates the programming code of the cost function
and provides exact values of the derivatives. In contrast to numerical differentiation it is not suffering from approximation errors and compared to symbolic differentiation it is more convenient to use, because no closed analytic expressions are required. Furthermore, we demonstrate how to generalize the parameter estimation scheme to delay differential equations, where estimating the delay time requires attention.
Features of the Jacobian matrix of the delay coordinates map are exploited for quantifying the robustness and reliability of state and parameter estimations for a given dynamical model using an observed time series. Relevant concepts of this approach
are introduced and illustrated for discrete and continuous time systems employing a filtered Henon map and a Rossler system.
Observability of state variables and parameters of a dynamical system from an observed time series is analyzed and quantified by means of the Jacobian matrix of the delay coordinates map. For each state variable and each parameter to be estimated a m
easure of uncertainty is introduced depending on the current state and parameter values, which allows us to identify regions in state and parameter space where the specific unknown quantity can (not) be estimated from a given time series. The method is demonstrated using the Ikeda map and the Hindmarsh-Rose model.
In this paper we describe a general probabilistic framework for modeling waveforms such as heartbeats from ECG data. The model is based on segmental hidden Markov models (as used in speech recognition) with the addition of random effects to the gener
ative model. The random effects component of the model handles shape variability across different waveforms within a general class of waveforms of similar shape. We show that this probabilistic model provides a unified framework for learning these models from sets of waveform data as well as parsing, classification, and prediction of new waveforms. We derive a computationally efficient EM algorithm to fit the model on multiple waveforms, and introduce a scoring method that evaluates a test waveform based on its shape. Results on two real-world data sets demonstrate that the random effects methodology leads to improved accuracy (compared to alternative approaches) on classification and segmentation of real-world waveforms.
