This python module uses statistical variational inference for predicting the evolution of physical dynamical systems. The package generates C++ files required for the solution of state and parameter estimation problems in the open-source large-scale nonlinear optimization software IPOPT. Equations of motion of user-defined dynamical systems are temporally discretized, and the Jacobian and Hessian matrices of the discrete system are calculated with symbolic differentiation. The equations of motion are imposed as strong constraints on the minimization of a cost function defining the distance between the model and time series observations of the system to be estimated. This code requires the installation of the following:
- Sympy (.py)
- Numpy (.py)
- IPOPT
IPOPT can be downloaded from https://projects.coin-or.org/Ipopt This software package was used in the following publications (in review):
[1] J. Taylor et al., Stochasticity and convergence in data
assimilation of predictive neuron models
[2] K.Abu-Hassan et al., Construction of neuromorphic models
of respiratory neurons