episode: Estimation with Penalisation In Systems of Ordinary...

Description Specify your ODE Specify loss function Optimise the loss function

Description

This package provide the tools for approximate and exact parameter estimation in ODE models with regularisation via penalisation.

Specify your ODE

The ode system is specified via the ode-subclasses: mak, plk, rlk and ratmak. In creating these you can also specify numerical solver type via solver and regularisation type of the parameter arguments via reg. To numerically solve the ODE use numsolve and to evaluate the ODE field use field. The differentials of both quantities can also be evaluated.

Specify loss function

To specify the loss function use the reg in the ODE object to control regularisation and opt to control the observations, their weights and the tuning parameter of the regularisation.

Optimise the loss function

Having an ode object and an opt object, there are two methods for estimating the parameters: approximate estimation via inverse collocation methods, aim, or exact estimation via interior point methods, rodeo. If desired, call rodeo on the results from aim to use the approximate estimates for initialising the exact estimation.


episode documentation built on May 1, 2019, 11:17 p.m.