Skip to content

Add additional Jacobian tests (including sparsity) #1211

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Apr 7, 2025
Merged

Add additional Jacobian tests (including sparsity) #1211

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
0079e96
init
TorkelE Apr 2, 2025
dc72438
add jacobian tests with simulations
TorkelE Apr 2, 2025
3b0a815
test fix
TorkelE Apr 3, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
111 changes: 107 additions & 4 deletions test/simulation_and_solving/jacobian_construction.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
### Prepares Tests ###

# Fetch packages.
using Catalyst, DiffEqBase, Test
using Catalyst, DiffEqBase, OrdinaryDiffEqRosenbrock, Test

# Sets stable rng number.
using StableRNGs
Expand All @@ -19,7 +19,7 @@ let
(3.0, 1.0), ∅ ↔ Y
(5.0, 2.0), X + Y ↔ XY
end

test_jac = jac_eval(jacobian_network_1, [:X => 1.0, :Y => 1.0, :XY => 1.0], [], 0.0)
real_jac = [-6.0 -5.0 2.0; -5.0 -6.0 2.0; 5.0 5.0 -2.0]
@test test_jac == real_jac
Expand All @@ -33,7 +33,7 @@ let
(p3 * X, 1.0), X + Y ↔ XY
end
@unpack X, Y, XY, p1, p2, p3 = jacobian_network_2

for factor in [1e-2, 1e-1, 1e0, 1e1, 1e2], repeat in 1:10
u = Dict(rnd_u0(jacobian_network_2, rng; factor))
p = Dict(rnd_ps(jacobian_network_2, rng; factor))
Expand Down Expand Up @@ -67,4 +67,107 @@ let
p[k3]*u[B] p[k3]*u[A] -p[k4]-3 * p[k5] * u[C]^2 / 2]
@test test_jac ≈ real_jac
end
end
end

# Checks that the Jacobians (dense and sparse) are identical for different system types.
Copy link
Member

@isaacsas isaacsas Apr 3, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do you ever test against a custom hand-coded Jacobian we know is right? That would be good to have in at least one case.

Copy link
Member Author

@TorkelE TorkelE Apr 3, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes we do (right now that's basically the only Jacobian tests we do actually): https://github.com/SciML/Catalyst.jl/blob/master/test/simulation_and_solving/jacobian_construction.jl

However, we do not do this for sparse though.

Copy link
Member

@isaacsas isaacsas Apr 3, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We should check sparse against the exact Jacobian, but I don't think we need an "exact" Jacobian that returns a sparse matrix (i.e. checking the Catalyst-generated sparse vs. a hand dense is fine).

let
# Creates model (vaguely messy model without conserved quantities).
rn = @reaction_network begin
(p,d), 0 <--> (X,Y)
(k1,k2), X + Y <--> XY
(k1,k2), X + 2Y <--> XY2
(k1,k2), XY + XY2 <--> X2Y3
d, (XY2,X2Y3) --> 0
mm(X2Y3,v,K), 0 --> Z
(k3,k4), 3Z <--> Z3
1.0, X3 --> 0
end

# Performs tests for different randomised values (to be thoroughly sure).
for factor in [0.1, 1.0, 10.0]
# Creates randomised species and parameter values. Generates jacobians (dense/sparse).
u0 = rnd_u0(rn, rng; factor)
ps = rnd_ps(rn, rng; factor)
oprob_jac = ODEProblem(rn, u0, 1.0, ps; jac = true, sparse = false)
oprob_sjac = ODEProblem(rn, u0, 1.0, ps; jac = true, sparse = true)
sprob_jac = SDEProblem(rn, u0, 1.0, ps; jac = true, sparse = false)
sprob_sjac = SDEProblem(rn, u0, 1.0, ps; jac = true, sparse = true)
nlprob_jac = NonlinearProblem(rn, u0, ps; jac = true, sparse = false)
nlprob_sjac = NonlinearProblem(rn, u0, ps; jac = true, sparse = true)

# Checks that Jacobians ar identical.
function eval_jac(prob, sparse)
J = sparse ? deepcopy(prob.f.jac_prototype) : zeros(length(prob.u0), length(prob.u0))
ModelingToolkit.is_time_dependent(prob) ? prob.f.jac(J, prob.u0, prob.p, 0.0) : prob.f.jac(J, prob.u0, prob.p)
return J
end
@test eval_jac(oprob_jac, false) == eval_jac(sprob_jac, false) == eval_jac(nlprob_jac, false)
@test_broken eval_jac(oprob_sjac, true) == eval_jac(sprob_sjac, true) == eval_jac(nlprob_sjac, true) # https://github.com/SciML/ModelingToolkit.jl/issues/3527
end
end

### Sparse Jacobian Tests ###

# Checks that generated dense/sparse Jacobians are identical.
let
# Creates model (vaguely messy model without conserved quantities).
# Model includes a time-dependent reaction.
rn = @reaction_network begin
(p,d), 0 <--> (X,Y,Z)
k1, X + Y --> XY
k2, X + 2Z --> XZ2
k3, Y3 +X2 --> Y3Z2
k4, X + Y + Z --> XYZ
k5, XZ2 + Y3Z2 --> XY3Z4
k6, XYZ + XYZ --> X2Y2Z2
d, (XY3Z4, X2Y2Z2) --> 0
X + Y, V --> 0
k7/(1 + t), 2V --> V2
Z, V2 --> 0
end

# Performs tests for different randomised values (to be thoroughly sure).
for factor in [0.1, 1.0, 10.0]
# Creates randomised species and parameter values. Generates jacobians (dense/sparse).
u0 = rnd_u0(rn, rng; factor)
t_val = factor*rand()
ps = rnd_ps(rn, rng; factor)
jac = jac_eval(rn, u0, ps, t_val; sparse = false)
jac_sparse = jac_eval(rn, u0, ps, t_val; sparse = true)

# Check correctness (both by converting to sparse jac to dense, and through multiplication with other matrix).
@test Matrix(jac_sparse) == jac
mat = factor*rand(rng, length(u0), length(u0))
@test jac_sparse * mat ≈ jac * mat
end
end

# Tests that simulations with different Jacobian and sparsity options are identical.
let
# Creates model (vaguely messy model without conserved quantities).
rn = @reaction_network begin
(v0 + mm(X,v,K),d), 0 <--> X + 2Y
(k1,k2), X + Y <--> XY
(k1,k2), X + Y2 <--> XY2
(k3,k4), XY + XY2 <--> X2Y3
1.0, (XY,XY2,X2Y3) --> 0
mm(X2Y3,v,K), 0 --> Z
(k3*X,k4*Y), 3Z <--> Z3
d, Z --> 0
end

# Generates initial conditions and parameter values. Creates problems with/o (sparse/dense) jacobian.
u0 = rnd_u0(rn, rng)
ps = rnd_ps(rn, rng)
oprob = ODEProblem(rn, u0, 1.0, ps; jac = false, sparse = false)
oprob_j = ODEProblem(rn, u0, 1.0, ps; jac = true, sparse = false)
oprob_s = ODEProblem(rn, u0, 1.0, ps; jac = false, sparse = true)
oprob_js = ODEProblem(rn, u0, 1.0, ps; jac = true, sparse = true)

# Simulates system with implicit solver. Checks that all solutions are identical.
sol = solve(oprob, Rosenbrock23(), saveat = 0.1, abstol = 1e-8, reltol = 1e-8)
sol_j = solve(oprob_j, Rosenbrock23(), saveat = 0.1, abstol = 1e-8, reltol = 1e-8)
sol_s = solve(oprob_s, Rosenbrock23(), saveat = 0.1, abstol = 1e-8, reltol = 1e-8)
sol_js = solve(oprob_js, Rosenbrock23(), saveat = 0.1, abstol = 1e-8, reltol = 1e-8)
@test sol ≈ sol_j ≈ sol_s ≈ sol_js
end
12 changes: 6 additions & 6 deletions test/test_functions.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ end
# Evaluates the the drift function of the ODE corresponding to a reaction network.
# Also checks that in place and out of place evaluations are identical.
function f_eval(rs::ReactionSystem, u, p, t; combinatoric_ratelaws = true)
prob = ODEProblem(rs, u, (0.0, 0.0), p; combinatoric_ratelaws)
prob = ODEProblem(rs, u, 0.0, p; combinatoric_ratelaws)
du = zeros(length(u))
prob.f(du, prob.u0, prob.p, t)
@test du == prob.f(prob.u0, prob.p, t)
Expand All @@ -47,9 +47,9 @@ end

# Evaluates the the Jacobian of the drift function of the ODE corresponding to a reaction network.
# Also checks that in place and out of place evaluations are identical.
function jac_eval(rs::ReactionSystem, u, p, t; combinatoric_ratelaws = true)
prob = ODEProblem(rs, u, (0.0, 0.0), p; jac = true, combinatoric_ratelaws)
J = zeros(length(u), length(u))
function jac_eval(rs::ReactionSystem, u, p, t; combinatoric_ratelaws = true, sparse = false)
prob = ODEProblem(rs, u, 0.0, p; jac = true, combinatoric_ratelaws, sparse)
J = sparse ? deepcopy(prob.f.jac_prototype) : zeros(length(u), length(u))
prob.f.jac(J, prob.u0, prob.p, t)
@test J == prob.f.jac(prob.u0, prob.p, t)
return J
Expand All @@ -58,9 +58,9 @@ end
# Evaluates the the diffusion function of the SDE corresponding to a reaction network.
# Also checks that in place and out of place evaluations are identical.
function g_eval(rs::ReactionSystem, u, p, t; combinatoric_ratelaws = true)
prob = SDEProblem(rs, u, (0.0, 0.0), p; combinatoric_ratelaws)
prob = SDEProblem(rs, u, 0.0, p; combinatoric_ratelaws)
dW = zeros(length(u), numreactions(rs))
prob.g(dW, prob.u0, prob.p, t)
@test dW == prob.g(prob.u0, prob.p, t)
return dW
end
end
Loading