Expectation Algorithms

MonteCarlo(trajectories::Int)

Estimate expectations by averaging random samples from the uncertainty distribution.

Arguments

• trajectories: Number of samples or ensemble trajectories to use.

Fields

• trajectories: Stored sample or trajectory count.

Returns

A MonteCarlo expectation algorithm for solve.

Koopman()
Koopman(sensealg::AbstractExpectationADAlgorithm)

Solve expectations by quadrature over the uncertainty domain.

Koopman builds an Integrals.jl integral problem for the observable and distribution in an ExpectationProblem. The optional sensealg selects the automatic differentiation strategy used by differentiable expectation solves.

Arguments

• sensealg: Automatic differentiation strategy. Defaults to NonfusedAD().

Fields

• sensealg: Stored automatic differentiation strategy.

Returns

A Koopman expectation algorithm for solve.

NonfusedAD()

Use separate quadrature solves for the primal integral and parameter pullback in Koopman expectation solves.

Returns

A NonfusedAD automatic differentiation algorithm for Koopman.

PrefusedAD()
PrefusedAD(norm_partials::Bool)

Fuse the primal integrand and parameter partials before the pullback in Koopman expectation solves.

Arguments

• norm_partials: Whether fused quadrature norms should include partial terms in implementations that use this option. Defaults to true.

Fields

• norm_partials: Stored partial-norm flag.

Returns

A PrefusedAD automatic differentiation algorithm for Koopman.

PostfusedAD()
PostfusedAD(norm_partials::Bool)

Fuse the primal integrand and parameter partials during the pullback after the primal Koopman expectation solve.

Arguments

• norm_partials: Whether fused quadrature norms should include partial terms in implementations that use this option. Defaults to true.

Fields

• norm_partials: Stored partial-norm flag.

Returns

A PostfusedAD automatic differentiation algorithm for Koopman.