petsc-3.11.2 2019-05-18
Report Typos and Errors

Sensitivity analysis for ODE and DAE: Examples

The TS library provides discrete adjoint models (TSAdjoint) and tangent linear models (TSForward) for sensitivity analysis for ODEs and DAEs. The adjoint solvers support gradient calculation for multiple cost functions, and the tangent linear solvers support gradient calculation with respect to multiple parameters. Adjoint is particularly efficient when the number of cost functions is much less than the number of parameters. Note that the parameters can be initial states or system parameters as used in the calculation of ODE/DAE right-hand sides. Typical cost functions of interest may depend on the final solution to the ODE/DAE or on the whole trajectory (taking an integral form). The integral can be evaluated together with the time integration.

Beginner - Basic usage
TSForwardSetSensitivities TSSetCostGradients
Intermediate - Setting options for algorithms and data structures
TSAdjointMonitorCancel TSAdjointSetSteps TSForwardSetIntegralGradients
TSAdjointMonitorDefault TSAdjointSolve TSGetCostGradients
TSAdjointMonitorDrawSensi TSAdjointStep TSGetCostIntegral
TSAdjointMonitorSensi TSForwardGetIntegralGradients TSSetCostIntegrand
TSAdjointMonitorSet TSForwardGetSensitivities TSSetRHSJacobianP
Advanced - Setting more advanced options and customization
TSAdjointCostIntegral TSForwardCostIntegral TSForwardStep
TSAdjointSetUp TSForwardSetUp
Developer - Interfaces intended primarily for library developers, not for typical applications programmers
TSAdjointMonitor TSComputeCostIntegrand TSComputeDRDYFunction
TSAdjointMonitorSetFromOptions TSComputeDRDPFunction TSComputeRHSJacobianP
Deprecated - Functionality scheduled for removal in future versions
TSAdjointComputeDRDPFunction TSAdjointComputeRHSJacobian
TSAdjointComputeDRDYFunction TSAdjointSetRHSJacobian

Table of Contents