Skip to content

GitLab

Commit 6c4488ce authored by Michel Juillard's avatar Michel Juillard
Browse files

fixed computation of log likelihood and kalman_filter

parent f02dc2fc
Hide whitespace changes
Inline Side-by-side
src/KalmanFilterTools.jl
View file @ 6c4488ce
......@@ -25,7 +25,8 @@ using LinearAlgebra.BLAS
export KalmanLikelihoodWs, FastKalmanLikelihoodWs, DiffuseKalmanLikelihoodWs
export DiffuseKalmanFilterWs
export KalmanSmootherWs, kalman_likelihood, kalman_likelihood_monitored
export fast_kalman_likelihood, diffuse_kalman_likelihood, kalman_filter!
export fast_kalman_likelihood, diffuse_kalman_likelihood
export kalman_filter!, diffuse_kalman_filter!
export kalman_smoother!
abstract type KalmanWs{T, U} end
......
src/kalman_base.jl
View file @ 6c4488ce
......@@ -239,6 +239,11 @@ function get_v!(v::AbstractVector{T}, y::AbstractVecOrMat{T}, z::AbstractVector{
v .= v .- az
end
# v = y - a[z] -- Z selection matrix -- univariate
function get_v!(y::AbstractVecOrMat{T}, z::AbstractVector{U}, a::AbstractVector{T}, i::U) where {T <: AbstractFloat, U <: Integer}
return y[i] - a[z[i]]
end
# v = y - Z*a -- missing observations
function get_v!(v::AbstractVector{T}, y::AbstractVecOrMat{T}, z::AbstractMatrix{T}, a::AbstractVector{T}, t::U, pattern::Vector{U}) where {T <: AbstractFloat, U <: Integer}
v .= view(y, pattern, t)
......@@ -251,12 +256,21 @@ function get_v!(v::AbstractVector{T}, y::AbstractVecOrMat{T}, z::AbstractVector{
end
# v = y - c - Z*a -- basic
function get_v!(v::AbstractArray{T}, y::AbstractVecOrMat{T}, c::AbstractArray{T}, z::AbstractArray{T}, a::AbstractArray{T}, iy::U, ny::U) where {T <: AbstractFloat, U <: Integer}
function get_v!(v::AbstractArray{T}, y::AbstractVecOrMat{T}, c::AbstractVector{T}, z::AbstractArray{T}, a::AbstractArray{T}, iy::U, ny::U) where {T <: AbstractFloat, U <: Integer}
copyto!(v, 1, y, iy, ny)
v .-= c
gemm!('N', 'N', -1.0, z, a, 1.0, v)
end
# v = y - c - Z*a -- basic -- univariate
function get_v!(Y::AbstractVecOrMat{T}, c::AbstractVector{T}, Z::AbstractVecOrMat{T}, a::AbstractVector{T}, i::U) where {T <: AbstractFloat, U <: Integer}
v = Y[i] - c[i]
@inbounds @simd for j = 1:length(a)
v -= Z[i, j]*a[j]
end
return v
end
# v = y - c - a[z] -- Z selection matrix
function get_v!(v::AbstractArray{T}, y::AbstractVecOrMat{T}, c::AbstractArray{T}, z::AbstractVector{U}, a::AbstractArray{T}, iy::U, ny::U) where {T <: AbstractFloat, U <: Integer}
copyto!(v, 1, y, iy, ny)
......@@ -264,6 +278,11 @@ function get_v!(v::AbstractArray{T}, y::AbstractVecOrMat{T}, c::AbstractArray{T}
v .-= c .+ az
end
# v = y - c - a[z] -- Z selection matrix -- univariate
function get_v!(y::AbstractVecOrMat{T}, c::AbstractVector{T}, z::AbstractVector{U}, a::AbstractVector{T}, i::U) where {T <: AbstractFloat, U <: Integer}
return y[i] - c[i] - a[z[i]]
end
# v = y - c - Z*a -- missing observations
function get_v!(v::AbstractArray{T}, y::AbstractVecOrMat{T}, c::AbstractArray{T}, z::AbstractArray{T}, a::AbstractArray{T}, t::U, pattern::Vector{U}) where {T <: AbstractFloat, U <: Integer}
v .= view(y, pattern, t) .- view(c, pattern)
......@@ -298,6 +317,26 @@ function get_Veta!(Veta, Q, R, N, RQ, tmp)
end
function get_vZsmall(Zsmall::AbstractMatrix{T}, iZsmall::AbstractVector{U}, Z::AbstractArray{T}, pattern::AbstractVector{U}, n::U, ny::U) where {T <: AbstractFloat, U <: Integer}
vZsmall = view(Zsmall, 1:n, :)
if n == ny
copyto!(vZsmall, Z)
else
vZsmall .= view(Z, pattern, :)
end
return vZsmall
end
function get_vZsmall(Zsmall::AbstractMatrix{T}, iZsmall::AbstractVector{U}, z::AbstractArray{U}, pattern::AbstractVector{U}, n::U, ny::U) where {T <: AbstractFloat, U <: Integer}
vZsmall = view(iZsmall, 1:n)
if n == ny
copyto!(vZsmall, z)
else
vZsmall .= view(z, pattern)
end
return vZsmall
end
function get_vZsmall(Zsmall::AbstractMatrix{T}, iZsmall::AbstractVector{U}, Z::AbstractArray{T}, pattern::AbstractVector{U}, n::U, ny::U, t::U) where {T <: AbstractFloat, U <: Integer}
changeZ = ndims(Z) > 2
vZ = changeZ ? view(Z, :, :, t) : view(Z, :, :)
vZsmall = view(Zsmall, 1:n, :)
......@@ -309,7 +348,7 @@ function get_vZsmall(Zsmall::AbstractMatrix{T}, iZsmall::AbstractVector{U}, Z::A
return vZsmall
end
function get_vZsmall(Zsmall::AbstractMatrix{T}, iZsmall::AbstractVector{U}, z::AbstractArray{U}, pattern::AbstractVector{U}, n::U, ny::U) where {T <: AbstractFloat, U <: Integer}
function get_vZsmall(Zsmall::AbstractMatrix{T}, iZsmall::AbstractVector{U}, z::AbstractArray{U}, pattern::AbstractVector{U}, n::U, ny::U, t::U) where {T <: AbstractFloat, U <: Integer}
changeZ = ndims(z) > 1
vz = changeZ ? view(z, :, t) : view(z, :,)
vZsmall = view(iZsmall, 1:n)
......
src/kalman_filter.jl
View file @ 6c4488ce
......@@ -33,13 +33,12 @@ function kalman_filter!(Y::AbstractArray{U},
vQ = view(Q, :, :, 1)
get_QQ!(ws.QQ, vR, vQ, ws.RQ)
l2pi = log(2*pi)
fill!(ws.lik, 0.0)
t = start
steady = false
vP = view(P, :, :, 1)
copy!(ws.oldP, vP)
cholHset = false
@inbounds while t <= last
while t <= last
pattern = data_pattern[t]
ndata = length(pattern)
......@@ -79,7 +78,7 @@ function kalman_filter!(Y::AbstractArray{U},
get_cholF!(ws.cholH, vvH)
cholHset = true
end
ws.lik[t] = univariate_step!(Y, t, ws.Zsmall, vvH, T, ws.QQ, va, vP, ws.kalman_tol, ws)
ws.lik[t] = ndata*l2pi + univariate_step!(Y, t, c, ws.Zsmall, vvH, d, T, ws.QQ, va, vP, ws.kalman_tol, ws, pattern)
t += 1
continue
end
......@@ -108,12 +107,8 @@ function kalman_filter!(Y::AbstractArray{U},
end
t += 1
end
@inbounds if presample > 0
LIK = -0.5*(sum(view(ws.lik,(presample+1):nobs)))
else
LIK = -0.5*(sum(ws.lik))
end
LIK
vlik = view(ws.lik, start + presample:last)
return -0.5*sum(vlik)
end
struct DiffuseKalmanFilterWs{T, U} <: KalmanWs{T, U}
......@@ -135,8 +130,9 @@ struct DiffuseKalmanFilterWs{T, U} <: KalmanWs{T, U}
ZPstar::Matrix{T}
Kinf::Matrix{T}
iFZ::Matrix{T}
Kstar::Matrix{T}
K::Matrix{T}
PTmp::Matrix{T}
oldP::Matrix{T}
uKinf::Vector{T}
uKstar::Vector{T}
Kinf_Finf::Vector{T}
......@@ -166,8 +162,9 @@ struct DiffuseKalmanFilterWs{T, U} <: KalmanWs{T, U}
ZPstar = Matrix{T}(undef, ny, ns)
Kinf = Matrix{T}(undef, ny, ns)
iFZ = Matrix{T}(undef, ny, ns)
Kstar = Matrix{T}(undef, ny, ns)
K = Matrix{T}(undef, ny, ns)
PTmp = Matrix{T}(undef, ns, ns)
oldP = Matrix{T}(undef, ns, ns)
uKinf = Vector{T}(undef, ns)
uKstar = Vector{T}(undef, ns)
Kinf_Finf = Vector{T}(undef, ns)
......@@ -178,7 +175,7 @@ struct DiffuseKalmanFilterWs{T, U} <: KalmanWs{T, U}
lik = zeros(T, nobs)
kalman_tol = 1e-12
new(csmall, Zsmall, iZsmall, QQ, RQ, c, v, F, iF, iFv, a1, cholF, cholH, ZP, Fstar,
ZPstar, Kinf, iFZ, Kstar, PTmp, uKinf, uKstar, Kinf_Finf, ystar, Zstar, Hstar,
ZPstar, Kinf, iFZ, K, PTmp, oldP, uKinf, uKstar, Kinf_Finf, ystar, Zstar, Hstar,
PZi, lik, kalman_tol)
end
end
......@@ -220,6 +217,7 @@ function diffuse_kalman_filter_init!(Y::AbstractArray{U},
vR = view(R, :, :, 1)
vQ = view(Q, :, :, 1)
get_QQ!(ws.QQ, vR, vQ, ws.RQ)
l2pi = log(2*pi)
diffuse_kalman_tol = 1e-8
kalman_tol = 1e-8
cholHset = false
......@@ -233,6 +231,9 @@ function diffuse_kalman_filter_init!(Y::AbstractArray{U},
vT = changeT ? view(T, :, :, t) : view(T, :, :)
vR = changeR ? view(R, :, :, t) : view(R, :, :)
vQ = changeQ ? view(Q, :, :, t) : view(Q, :, :)
if changeR || changeQ
get_QQ!(ws.QQ, vR, vQ, ws.RQ)
end
va = changeA ? view(a, :, t) : view(a, :)
va1 = changeA ? view(a, :, t + 1) : view(a, :)
vd = changeD ? view(d, :, t) : view(d, :)
......@@ -240,9 +241,6 @@ function diffuse_kalman_filter_init!(Y::AbstractArray{U},
vPinf1 = changePinf ? view(Pinf, :, :, t + 1) : view(Pinf, :, :)
vPstar = changePstar ? view(Pstar, :, :, t) : view(Pstar, :, :)
vPstar1 = changePstar ? view(Pstar, :, :, t + 1) : view(Pstar, :, :)
if changeR || changeQ
get_QQ!(ws.QQ, vR, vQ, ws.RQ)
end
vv = view(ws.v, 1:ndata)
vvH = view(vH, pattern, pattern)
vZP = view(ws.ZP, 1:ndata, :)
......@@ -256,7 +254,7 @@ function diffuse_kalman_filter_init!(Y::AbstractArray{U},
vcholH = view(ws.cholH, 1:ndata, 1:ndata, 1)
viFv = view(ws.iFv, 1:ndata)
vKinf = view(ws.Kinf, 1:ndata, :, 1)
vKstar = view(ws.Kstar, 1:ndata, :, 1)
vKstar = view(ws.K, 1:ndata, :, 1)
# v = Y[:,t] - c - Z*a
get_v!(vv, Y, vc, vZsmall, va, t, pattern)
......@@ -271,10 +269,10 @@ function diffuse_kalman_filter_init!(Y::AbstractArray{U},
get_cholF!(vcholH, vvH)
cholHset = true
end
ws.lik[t] += univariate_step(Y, t, vZsmall, vvH, T, QQ, a, Pinf, Pstar, diffuse_kalman_tol, kalman_tol, ws)
ws.lik[t] += ndata*l2pi + univariate_step(Y, t, c, vZsmall, vvH, d, T, QQ, a, Pinf, Pstar, diffuse_kalman_tol, kalman_tol, ws, pattern)
end
else
ws.lik[t] = log(det_from_cholesky(vcholF))
ws.lik[t] = ndata*l2pi + log(det_from_cholesky(vcholF))
# Kinf = iFinf*Z*Pinf %define Kinf'=T^{-1}*K_0 with M_{\infty}=Pinf*Z'
copy!(vKinf, vZP)
LAPACK.potrs!('U', vcholF, vKinf)
......@@ -296,66 +294,55 @@ function diffuse_kalman_filter_init!(Y::AbstractArray{U},
t
end
function diffuse_kalman_filter(Y::Matrix{U},
Z::AbstractArray{W},
H::Matrix{U},
T::Matrix{U},
R::Matrix{U},
Q::Matrix{U},
a::Vector{U},
Pinf::Matrix{U},
Pstar::Matrix{U},
start::V,
last::V,
presample::V,
tol::U,
ws::DiffuseKalmanLikelihoodWs) where {U <: AbstractFloat,
V <: Integer,
W <: Real}
function diffuse_kalman_filter!(Y::AbstractArray{U},
c::AbstractArray{U},
Z::AbstractArray{W},
H::AbstractArray{U},
d::AbstractArray{U},
T::AbstractArray{U},
R::AbstractArray{U},
Q::AbstractArray{U},
a::AbstractArray{U},
Pinf::AbstractArray{U},
Pstar::AbstractArray{U},
start::V,
last::V,
presample::V,
tol::U,
ws::DiffuseKalmanFilterWs,
data_pattern::Vector{Vector{V}}) where {U <: AbstractFloat,
V <: Integer,
W <: Real}
ny = size(Y,1)
nobs = last - start + 1
get_QQ!(ws.QQ, R, Q, ws.RQ)
lik_cst = (nobs - presample)*ny*log(2*pi)
fill!(ws.lik, 0.0)
t = diffuse_kalman_likelihood_init!(Y, Z, H, T, ws.QQ, a, Pinf, Pstar, start, last, tol, ws)
kalman_likelihood(Y, Z, H, T, R, Q, a, Pstar, t, last, presample, ws)
@inbounds if presample > 0
LIK = -0.5*(lik_cst + sum(view(ws.lik, (presample+1):nobs)))
else
LIK = -0.5*(lik_cst + sum(ws.lik))
end
LIK
t = diffuse_kalman_filter_init!(Y, c, Z, H, d, T, R, Q, a, Pinf, Pstar, start, last, presample, tol, ws, data_pattern)
kalman_filter!(Y, c, Z, H, d, T, R, Q, a, Pstar, t + 1, last, presample, ws, data_pattern)
vlik = view(ws.lik, start + presample:last)
return -0.5*sum(vlik)
end
function diffuse_kalman_filter!(Y::Matrix{U},
Z::AbstractArray{W},
H::Matrix{U},
T::Matrix{U},
R::Matrix{U},
Q::Matrix{U},
a::Vector{U},
Pinf::Matrix{U},
Pstar::Matrix{U},
start::V,
last::V,
presample::V,
tol::U,
ws::DiffuseKalmanFilterWs,
data_pattern::Vector{Vector{V}}) where {U <: AbstractFloat,
V <: Integer,
W <: Real}
ny = size(Y,1)
nobs = last - start + 1
get_QQ!(ws.QQ, R, Q, ws.RQ)
lik_cst = (nobs - presample)*ny*log(2*pi)
fill!(ws.lik, 0.0)
t = diffuse_kalman_filter_init!(Y, Z, H, T, ws.QQ, a, Pinf, Pstar, start, last, tol, ws, data_pattern)
kalman_likelihood(Y, Z, H, T, R, Q, a, Pstar, t, last, presample, ws, data_pattern)
@inbounds if presample > 0
LIK = -0.5*(lik_cst + sum(view(ws.lik, (presample+1):nobs)))
else
LIK = -0.5*(lik_cst + sum(ws.lik))
end
LIK
end
function diffuse_kalman_filter!(Y::AbstractArray{U},
c::AbstractArray{U},
Z::AbstractArray{W},
H::AbstractArray{U},
d::AbstractArray{U},
T::AbstractArray{U},
R::AbstractArray{U},
Q::AbstractArray{U},
a::AbstractArray{U},
Pinf::AbstractArray{U},
Pstar::AbstractArray{U},
start::V,
last::V,
presample::V,
tol::U,
ws::DiffuseKalmanFilterWs) where {U <: AbstractFloat,
V <: Integer,
W <: Real}
m, n = size(Y)
full_data_pattern = [collect(1:m) for i = 1:n]
diffuse_kalman_filter!(Y, c, Z, H, d, T, R, Q, a, Pinf, Pstar, start, last, presample, tol, ws, full_data_pattern)
end
src/kalman_likelihood.jl
View file @ 6c4488ce
......@@ -78,8 +78,6 @@ function kalman_likelihood(Y::AbstractArray{U},
nobs = last - start + 1
# QQ = R*Q*R'
get_QQ!(ws.QQ, R, Q, ws.RQ)
lik_cst = (nobs - presample)*ny*log(2*pi)
fill!(ws.lik, 0.0)
cholHset = false
t = start
iy = (start - 1)*ny + 1
......@@ -110,12 +108,9 @@ function kalman_likelihood(Y::AbstractArray{U},
end
t += 1
end
@inbounds if presample > 0
LIK = -0.5*(lik_cst + sum(view(ws.lik,(presample+1):nobs)))
else
LIK = -0.5*(lik_cst + sum(ws.lik))
end
LIK
lik_cst = (nobs - presample)*ny*log(2*pi)
vlik = view(ws.lik, start + presample:last)
return -0.5*(lik_cst + sum(vlik))
end
function kalman_likelihood(Y::AbstractArray{U},
......@@ -136,7 +131,6 @@ function kalman_likelihood(Y::AbstractArray{U},
# QQ = R*Q*R'
get_QQ!(ws.QQ, R, Q, ws.RQ)
l2pi = log(2*pi)
fill!(ws.lik, 0.0)
t = start
iy = (start - 1)*ny + 1
ncolZ = size(Z, 2)
......@@ -161,13 +155,12 @@ function kalman_likelihood(Y::AbstractArray{U},
get_F!(vF, vZP, vZsmall, P, vH)
info = get_cholF!(vcholF, vF)
if info != 0
@show info
# F is near singular
if !cholHset
get_cholF!(vcholH, vH)
cholHset = true
end
ws.lik[t] = univariate_step!(Y, t, Z, H, T, ws.QQ, a, P, ws.kalman_tol, ws)
ws.lik[t] = ndata*l2pi + univariate_step!(Y, t, Z, H, T, ws.QQ, a, P, ws.kalman_tol, ws, pattern)
else
# iFv = inv(F)*v
get_iFv!(viFv, vcholF, vv)
......@@ -181,12 +174,8 @@ function kalman_likelihood(Y::AbstractArray{U},
end
t += 1
end
@inbounds if presample > 0
LIK = -0.5*sum(view(ws.lik,(presample+1):nobs))
else
LIK = -0.5*sum(ws.lik)
end
LIK
vlik = view(ws.lik, start + presample:last)
return -0.5*sum(vlik)
end
function kalman_likelihood_monitored(Y::Matrix{U},
......@@ -206,8 +195,6 @@ function kalman_likelihood_monitored(Y::Matrix{U},
ns = size(T,1)
# QQ = R*Q*R'
get_QQ!(ws.QQ, R, Q, ws.RQ)
lik_cst = (nobs - presample)*ny*log(2*pi)
fill!(ws.lik, 0.0)
t = start
iy = (start - 1)*ny + 1
steady = false
......@@ -221,7 +208,6 @@ function kalman_likelihood_monitored(Y::Matrix{U},
get_F!(ws.F, ws.ZP, Z, P, H)
info = get_cholF!(ws.cholF, ws.F)
if info != 0
@show info
# F is near singular
if !cholHset
get_cholF!(ws.cholH, H)
......@@ -253,12 +239,9 @@ function kalman_likelihood_monitored(Y::Matrix{U},
end
t += 1
end
@inbounds if presample > 0
LIK = -0.5*(lik_cst + sum(view(ws.lik,(presample+1):nobs)))
else
LIK = -0.5*(lik_cst + sum(ws.lik))
end
LIK
lik_cst = (nobs - presample)*ny*log(2*pi)
vlik = view(ws.lik, start + presample:last)
return -0.5*(lik_cst + sum(vlik))
end
function kalman_likelihood_monitored(Y::AbstractArray{U},
......@@ -280,7 +263,6 @@ function kalman_likelihood_monitored(Y::AbstractArray{U},
# QQ = R*Q*R'
get_QQ!(ws.QQ, R, Q, ws.RQ)
l2pi = log(2*pi)
fill!(ws.lik, 0.0)
t = start
iy = (start - 1)*ny + 1
steady = false
......@@ -310,7 +292,7 @@ function kalman_likelihood_monitored(Y::AbstractArray{U},
get_cholF!(ws.cholH, H)
cholHset = true
end
ws.lik[t] = univariate_step!(Y, t, Z, H, T, ws.QQ, a, P, ws.kalman_tol, ws)
ws.lik[t] = ndatat*l2pi + univariate_step!(Y, t, Z, H, T, ws.QQ, a, P, ws.kalman_tol, ws, pattern)
t += 1
continue
end
......@@ -336,12 +318,8 @@ function kalman_likelihood_monitored(Y::AbstractArray{U},
end
t += 1
end
@inbounds if presample > 0
LIK = -0.5*(sum(view(ws.lik,(presample+1):nobs)))
else
LIK = -0.5*(sum(ws.lik))
end
LIK
vlik = view(ws.lik, start + presample:last)
return -0.5*sum(vlik)
end
struct FastKalmanLikelihoodWs{T, U} <: KalmanWs{T, U}
......@@ -425,8 +403,6 @@ function fast_kalman_likelihood(Y::Matrix{U},
ny = size(Y,1)
nobs = last - start + 1
get_QQ!(ws.QQ, R, Q, ws.RQ)
lik_cst = (nobs - presample)*ny*log(2*pi)
fill!(ws.lik, 0.0)
# F = Z*P*Z' + H
get_F!(ws.F, ws.ZP, Z, P, H)
get_cholF!(ws.cholF, ws.F)
......@@ -460,12 +436,9 @@ function fast_kalman_likelihood(Y::Matrix{U},
update_W!(ws.W, ws.ZW, ws.cholF, T, ws.K, ws.iFZW, ws.KtiFZW)
t += 1
end
@inbounds if presample > 0
LIK = -0.5*(lik_cst + sum(view(ws.lik, (presample+1):nobs)))
else
LIK = -0.5*(lik_cst + sum(ws.lik))
end
LIK
lik_cst = (nobs - presample)*ny*log(2*pi)
vlik = view(ws.lik, start + presample:last)
return -0.5*(lik_cst + sum(vlik))
end
function fast_kalman_likelihood(Y::Matrix{U},
......@@ -485,7 +458,6 @@ function fast_kalman_likelihood(Y::Matrix{U},
nobs = last - start + 1
get_QQ!(ws.QQ, R, Q, ws.RQ)
l2pi = log(2*pi)
fill!(ws.lik, 0.0)
# F = Z*P*Z' + H
get_F!(ws.F, ws.ZP, Z, P, H)
get_cholF!(ws.cholF, ws.F)
......@@ -534,12 +506,8 @@ function fast_kalman_likelihood(Y::Matrix{U},
update_W!(vW, vZW, vcholF, T, vK, viFZW, vKtiFZW)
t += 1
end
@inbounds if presample > 0
LIK = -0.5*(sum(view(ws.lik, (presample+1):nobs)))
else
LIK = -0.5*(sum(ws.lik))
end
LIK
vlik = view(ws.lik, start + presample:last)
return -0.5*sum(vlik)
end
struct DiffuseKalmanLikelihoodWs{T, U} <: KalmanWs{T, U}
......@@ -686,7 +654,6 @@ function diffuse_kalman_likelihood_init!(Y::Matrix{U},
ny = size(Y, 1)
t = start
LIK = 0
iy = (start - 1)*ny + 1
diffuse_kalman_tol = 1e-8
kalman_tol = 1e-8
......@@ -715,7 +682,7 @@ function diffuse_kalman_likelihood_init!(Y::Matrix{U},
if norm(vF) < tol
return t - 1
else
ws.lik[t] += univariate_step(Y, t, vZsmall, H, T, QQ, a, Pinf, Pstar, diffuse_kalman_tol, kalman_tol, pattern, ws)
ws.lik[t] += ndata*l2pi + univariate_step(Y, t, vZsmall, H, T, QQ, a, Pinf, Pstar, diffuse_kalman_tol, kalman_tol, pattern, ws, pattern)
end
else
ws.lik[t] = log(det_from_cholesky(ws.cholF))
......@@ -765,16 +732,11 @@ function diffuse_kalman_likelihood(Y::Matrix{U},
ny = size(Y,1)
nobs = last - start + 1
get_QQ!(ws.QQ, R, Q, ws.RQ)
lik_cst = (nobs - presample)*ny*log(2*pi)
fill!(ws.lik, 0.0)
t = diffuse_kalman_likelihood_init!(Y, Z, H, T, ws.QQ, a, Pinf, Pstar, start, last, tol, ws)
kalman_likelihood(Y, Z, H, T, R, Q, a, Pstar, t, last, presample, ws)
@inbounds if presample > 0
LIK = -0.5*(lik_cst + sum(view(ws.lik, (presample+1):nobs)))
else
LIK = -0.5*(lik_cst + sum(ws.lik))
end
LIK
kalman_likelihood(Y, Z, H, T, R, Q, a, Pstar, t + 1, last, presample, ws)
lik_cst = (nobs - presample)*ny*log(2*pi)
vlik = view(ws.lik, start + presample:last)
return -0.5*(lik_cst + sum(vlik))
end
function diffuse_kalman_likelihood(Y::Matrix{U},
......@@ -797,15 +759,9 @@ function diffuse_kalman_likelihood(Y::Matrix{U},
ny = size(Y,1)
nobs = last - start + 1
get_QQ!(ws.QQ, R, Q, ws.RQ)
lik_cst = (nobs - presample)*ny*log(2*pi)
fill!(ws.lik, 0.0)
t = diffuse_kalman_likelihood_init!(Y, Z, H, T, ws.QQ, a, Pinf, Pstar, start, last, tol, ws, data_pattern)
kalman_likelihood(Y, Z, H, T, R, Q, a, Pstar, t, last, presample, ws, data_pattern)
@inbounds if presample > 0
LIK = -0.5*(lik_cst + sum(view(ws.lik, (presample+1):nobs)))
else
LIK = -0.5*(lik_cst + sum(ws.lik))
end
LIK
kalman_likelihood(Y, Z, H, T, R, Q, a, Pstar, t + 1, last, presample, ws, data_pattern)
vlik = view(ws.lik, start + presample:last)
return -0.5*sum(vlik)
end
src/univariate_step.jl
View file @ 6c4488ce
......@@ -54,6 +54,38 @@ function univariate_step!(Y, t, Z, H, T, RQR, a, P, kalman_tol, ws)
return llik + 2*log(detLTcholH)
end
function univariate_step!(Y, t, Z, H, T, RQR, a, P, kalman_tol, ws, pattern)
ny = size(Y,1)
detLTcholH = 1.0
if !isdiag(H)
detLTcholH = transformed_measurement!(ws.ystar, ws.Zstar, view(Y, :, t), Z, ws.cholH)
H = I(ny)
else
copy!(ws.ystar, view(Y, :, t))
copy!(ws.Zstar, Z)
end
l2pi = log(2*pi)
llik = 0.0
ndata = size(pattern)
for i=1:ndata
Zi = view(ws.Zstar, pattern[i], :)
v = get_v!(ws.ystar, ws.Zstar, a, pattern[i])
F = get_F(Zi, P, H[i,i], ws.PZi)
if abs(F) > kalman_tol
a .+= (v/F) .* ws.PZi
# P = P - PZi*PZi'/F
ger!(-1.0/F, ws.PZi, ws.PZi, P)
llik += ndata*l2pi + log(F) + v*v/F
end
end
mul!(ws.a1, T, a)
a .= ws.a1
mul!(ws.PTmp, T, P)
copy!(P, RQR)
mul!(P, ws.PTmp, T', 1.0, 1.0)
return llik + 2*log(detLTcholH)
end
function univariate_step(Y, t, Z, H, T, QQ, a, Pinf, Pstar, diffuse_kalman_tol, kalman_tol, ws)
ny = size(Y,1)
detLTcholH = 1.0
......@@ -67,6 +99,7 @@ function univariate_step(Y, t, Z, H, T, QQ, a, Pinf, Pstar, diffuse_kalman_tol,
llik = 0.0