Hacky fix for approx periodic (#122)

* Add failing test and fix formatting

* Write to_sde for approx periodic

* Bump patch

* Restrict usage to Arrays

* Remove known failure case

* Fix ambiguity

* Remove test for disallowed implementation

Project.toml

@@ -1,7 +1,7 @@
 name = "TemporalGPs"
 uuid = "e155a3c4-0841-43e1-8b83-a0e4f03cc18f"
 authors = ["willtebbutt <[email protected]> and contributors"]
-version = "0.6.5"
+version = "0.6.6"
 
 [deps]
 AbstractGPs = "99985d1d-32ba-4be9-9821-2ec096f28918"

src/gp/lti_sde.jl

@@ -256,7 +256,8 @@ function stationary_distribution(::CosineKernel, ::SArrayStorage{T}) where {T<:R
     return Gaussian(m, P)
 end
 
-# Approximate Periodic Kernel
+# ApproxPeriodicKernel
+
 # The periodic kernel is approximated by a sum of cosine kernels with different frequencies.
 struct ApproxPeriodicKernel{N,K<:PeriodicKernel} <: KernelFunctions.SimpleKernel
     kernel::K
@@ -279,53 +280,37 @@ function Base.show(io::IO, κ::ApproxPeriodicKernel{N}) where {N}
     return print(io, "Approximate Periodic Kernel, (r = $(only(κ.kernel.r))) approximated with $N cosine kernels")
 end
 
-function lgssm_components(approx::ApproxPeriodicKernel{N}, t::Union{StepRangeLen, RegularSpacing}, storage::StorageType{T}) where {N,T<:Real}
-    Fs, Hs, ms, Ps = _init_periodic_kernel_lgssm(approx.kernel, storage, N)
-    nt = length(t)
-    As = map(F -> Fill(time_exp(F, T(step(t))), nt), Fs)
-    return _reduce_sum_cosine_kernel_lgssm(As, Hs, ms, Ps, N, nt, T)
-end
-function lgssm_components(approx::ApproxPeriodicKernel{N}, t::AbstractVector{<:Real}, storage::StorageType{T}) where {N,T<:Real}
-    Fs, Hs, ms, Ps = _init_periodic_kernel_lgssm(approx.kernel, storage, N)
-    t = vcat([first(t) - 1], t)
-    nt = length(diff(t))
-    As = _map(F -> _map(Δt -> time_exp(F, T(Δt)), diff(t)), Fs)
-    return _reduce_sum_cosine_kernel_lgssm(As, Hs, ms, Ps, N, nt, T)
-end
+# Can't use approx periodic kernel with static arrays -- the dimensions become too large.
+_ap_error() = throw(error("Unable to construct an ApproxPeriodicKernel for SArrayStorage"))
+to_sde(::ApproxPeriodicKernel, ::SArrayStorage) = _ap_error()
+stationary_distribution(::ApproxPeriodicKernel, ::SArrayStorage) = _ap_error()
 
-function _init_periodic_kernel_lgssm(kernel::PeriodicKernel, storage, N::Int=7)
-    r = kernel.r
-    l⁻² = inv(4 * only(r)^2)
-
+function to_sde(::ApproxPeriodicKernel{N}, storage::ArrayStorage{T}) where {T<:Real, N}
+
+    # Compute F and H for component processes.
     F, _, H = to_sde(CosineKernel(), storage)
     Fs = ntuple(N) do i
         2π * (i - 1) * F
     end
-    Hs = Fill(H, N)
 
+    # Combine component processes into a single whole.
+    F = block_diagonal(collect.(Fs)...)
+    q = zero(T)
+    H = repeat(collect(H), N)
+    return F, q, H
+end
+
+function stationary_distribution(kernel::ApproxPeriodicKernel{N}, storage::ArrayStorage{<:Real}) where {N}
     x0 = stationary_distribution(CosineKernel(), storage)
-    ms = Fill(x0.m, N)
-    P = x0.P
+    m = collect(repeat(x0.m, N))
+    r = kernel.kernel.r
+    l⁻² = inv(4 * only(r)^2)
     Ps = ntuple(N) do j
         qⱼ = (1 + (j !== 1) ) * besseli(j - 1, l⁻²) / exp(l⁻²)
-        qⱼ * P
-    end    
-
-    Fs, Hs, ms, Ps
-end
-
-function _reduce_sum_cosine_kernel_lgssm(As, Hs, ms, Ps, N, nt, T)
-    as = Fill(Fill(Zeros{T}(size(first(first(As)), 1)), nt), N)
-    Qs = _map((P, A) -> _map(A -> Symmetric(P) - A * Symmetric(P) * A', A), Ps, As)
-    Hs = Fill(vcat(Hs...), nt)
-    h = Fill(zero(T), nt)
-    As = _map(block_diagonal, As...)
-    as = -map(vcat, as...)
-    Qs = _map(block_diagonal, Qs...)
-    m = reduce(vcat, ms)
-    P = block_diagonal(Ps...)
-    x0 = Gaussian(m, P)
-    return As, as, Qs, (Hs, h), x0
+        return qⱼ * x0.P
+    end
+    P = collect(block_diagonal(Ps...))
+    return Gaussian(m, P)
 end
 
 # Constant

test/gp/lti_sde.jl

@@ -29,7 +29,7 @@ end
     @testset "$(typeof(t)), $storage, $N" for t in (
             sort(rand(Nt)), RegularSpacing(0.0, 0.1, Nt)
         ),
-        storage in (SArrayStorage{Float64}(), ArrayStorage{Float64}()),
+        storage in (ArrayStorage{Float64}(), ),
         N in (5, 8)
 
         k = ApproxPeriodicKernel{N}()
@@ -131,6 +131,12 @@ println("lti_sde:")
                 val=3.0 * Matern32Kernel() * Matern52Kernel() * ConstantKernel(),
                 to_vec_grad=nothing,
             ),
+            # THIS IS KNOWN NOT TO WORK!
+            # (
+            #     name="prod-(Matern32Kernel + ConstantKernel) * Matern52Kernel",
+            #     val=(Matern32Kernel() + ConstantKernel()) * Matern52Kernel(),
+            #     to_vec_grad=nothing,
+            # ),
 
             # Summed kernels.
             (
@@ -149,18 +155,21 @@ println("lti_sde:")
         )
 
         # Construct a Gauss-Markov model with either dense storage or static storage.
-        storages = ((name="dense storage Float64", val=ArrayStorage(Float64)),
-        # (name="static storage Float64", val=SArrayStorage(Float64)),
-)
+        storages = (
+            (name="dense storage Float64", val=ArrayStorage(Float64)),
+            # (name="static storage Float64", val=SArrayStorage(Float64)),
+        )
 
         # Either regular spacing or irregular spacing in time.
-        ts = ((name="irregular spacing", val=collect(RegularSpacing(0.0, 0.3, N))),
-        # (name="regular spacing", val=RegularSpacing(0.0, 0.3, N)),
-)
+        ts = (
+            (name="irregular spacing", val=collect(RegularSpacing(0.0, 0.3, N))),
+            # (name="regular spacing", val=RegularSpacing(0.0, 0.3, N)),
+        )
 
-        σ²s = ((name="homoscedastic noise", val=(0.1,)),
-        # (name="heteroscedastic noise", val=(rand(rng, N) .+ 1e-1, )),
-)
+        σ²s = (
+            (name="homoscedastic noise", val=(0.1,)),
+            # (name="heteroscedastic noise", val=(rand(rng, N) .+ 1e-1, )),
+        )
 
         means = (
             (name="Zero Mean", val=ZeroMean()),