wupeifan

# Forward Kalman
function solve_kalman(m::AbstractFirstOrderExpectationalDifferenceModel, sol::FirstOrderSolution, Q, obs, Ω, x_0 = nothing)
    @unpack n, n_x, n_y, n_p, n_ϵ, η = m
    @unpack h_x, g_x, h_x_p, g_x_p, Σ, Σ_p = sol
    (isnothing(x_0) || length(x_0) == n_x) ||
        throw(ArgumentError("Length of x_0 mismatches model"))

    T = size(obs, 1)
    n_z = size(Q, 1)
    z = [zeros(n_z) for _ in 1:T]

wupeifan

using DifferentiableStateSpaceModels, ModelingToolkit, SparseArrays, LinearAlgebra, Parameters, Test, TimerOutputs, BenchmarkTools
using Turing, Zygote, ChainRules, Optim
Turing.setadbackend(:zygote)

# Generate fake data
H, mod_vals = Examples.rbc()
model = FirstOrderStateSpaceModel(H; mod_vals...)
p = [0.2, 0.02, 0.01]
θ_value = [0.5, 0.95]
solution, _ = solve_model!(model, θ = θ_value, p = p)

wupeifan

using DifferentiableStateSpaceModels, ModelingToolkit, SparseArrays, LinearAlgebra, Parameters, Test, TimerOutputs, BenchmarkTools
using Turing, Zygote, ChainRules
Turing.setadbackend(:zygote)

# Generate fake data
H, mod_vals = Examples.rbc()
model = FirstOrderStateSpaceModel(H; mod_vals...)
p = [0.2, 0.02, 0.01]
θ_value = [0.5, 0.95]
solution, _ = solve_model!(model, θ = θ_value, p = p)