Learn R Programming

deGradInfer (version 1.0.1)

Parameter Inference for Systems of Differential Equation

Description

Efficient Bayesian parameter inference for systems of ordinary differential equations. The inference is based on adaptive gradient matching (AGM, Dondelinger et al. 2013 , Macdonald 2017 ), which offers orders-of-magnitude improvements in computational efficiency over standard methods that require solving the differential equation system. Features of the package include flexible specification of custom ODE systems as R functions, support for missing variables, Bayesian inference via population MCMC.

Copy Link

Version

Install

install.packages('deGradInfer')

Monthly Downloads

7

Version

1.0.1

License

GPL-3

Maintainer

Frank Dondelinger

Last Published

January 20th, 2020

Functions in deGradInfer (1.0.1)

sigmoidVarKernParamInit

Auxiliary function for sigmoid kernel (used by gptk)
sigmoidVarKernGradient

Compute gradient of sigmoid kernel with respect to each parameter (used by gptk)
solveODE

Solve ODE system explicitly.
sigmoidVarKernExpandParam

Insert parameters into sigmoid kernel (used by gptk)
sigmoidVarKernDiagCompute

Compute diagonal of sigmoid kernel (used by gptk).
proposeParamsMCMC

Sample from proposal distribution for MCMC
sigmoidVarKernCompute

Compute K(x, x2) for sigmoid kernel, used by gptk
LV_example_dataset

Data from a Lotka-Volterra ODE system with 2 species and 4 parameters. Species in order are: 1. Sheep (Prey) 2. Wolves (Predators)
agm

Main function for adaptive gradient matching
sigmoidVarKernExtractParam

Auxiliary function for sigmoid kernel (used by gptk)
doMCMC

Main MCMC function Runs the MCMC for the specified number of iterations and returns the sampled parameter values
getODEGradient

Calculate gradients from ODE system