Merge pull request #987 from ArnoStrouwen/fix_docs

fix documentation examples
SciML · Jan 23, 2024 · b893d7a · b893d7a
2 parents aa4dc34 + 77ab69d
commit b893d7a
Show file tree

Hide file tree

Showing 13 changed files with 35 additions and 23 deletions.
diff --git a/docs/Project.toml b/docs/Project.toml
@@ -15,6 +15,7 @@ IterTools = "c8e1da08-722c-5040-9ed9-7db0dc04731e"
 Lux = "b2108857-7c20-44ae-9111-449ecde12c47"
 LuxCUDA = "d0bbae9a-e099-4d5b-a835-1c6931763bda"
 Optimization = "7f7a1694-90dd-40f0-9382-eb1efda571ba"
+OptimizationFlux = "253f991c-a7b2-45f8-8852-8b9a9df78a86"
 OptimizationNLopt = "4e6fcdb7-1186-4e1f-a706-475e75c168bb"
 OptimizationOptimJL = "36348300-93cb-4f02-beb5-3c3902f8871e"
 OptimizationOptimisers = "42dfb2eb-d2b4-4451-abcd-913932933ac1"
@@ -26,6 +27,7 @@ RecursiveArrayTools = "731186ca-8d62-57ce-b412-fbd966d074cd"
 ReverseDiff = "37e2e3b7-166d-5795-8a7a-e32c996b4267"
 SciMLSensitivity = "1ed8b502-d754-442c-8d5d-10ac956f44a1"
 SimpleChains = "de6bee2f-e2f4-4ec7-b6ed-219cc6f6e9e5"
+StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 StochasticDiffEq = "789caeaf-c7a9-5a7d-9973-96adeb23e2a0"
 Tracker = "9f7883ad-71c0-57eb-9f7f-b5c9e6d3789c"

diff --git a/docs/src/examples/dae/physical_constraints.md b/docs/src/examples/dae/physical_constraints.md
@@ -9,6 +9,7 @@ zeros, then we have a constraint defined by the right-hand side. Using
 terms must add to one. An example of this is as follows:
 
 ```@example dae
+using SciMLSensitivity
 using Lux, ComponentArrays, DiffEqFlux, Optimization, OptimizationNLopt,
     OrdinaryDiffEq, Plots
 
@@ -73,6 +74,7 @@ result_stiff = Optimization.solve(optprob, NLopt.LD_LBFGS(), maxiters = 100)
 ### Load Packages
 
 ```@example dae2
+using SciMLSensitivity
 using Lux, ComponentArrays, DiffEqFlux, Optimization, OptimizationNLopt,
     OrdinaryDiffEq, Plots
 

diff --git a/docs/src/examples/neural_ode/minibatch.md b/docs/src/examples/neural_ode/minibatch.md
@@ -1,6 +1,7 @@
 # Training a Neural Ordinary Differential Equation with Mini-Batching
 
 ```@example
+using SciMLSensitivity
 using DifferentialEquations, Flux, Random, Plots
 using IterTools: ncycle
 
@@ -95,6 +96,7 @@ When training a neural network, we need to find the gradient with respect to our
 For this example, we will use a very simple ordinary differential equation, newtons law of cooling. We can represent this in Julia like so.
 
 ```@example minibatch
+using SciMLSensitivity
 using DifferentialEquations, Flux, Random, Plots
 using IterTools: ncycle
 

diff --git a/docs/src/examples/neural_ode/neural_gde.md b/docs/src/examples/neural_ode/neural_gde.md
@@ -2,7 +2,7 @@
 
 This tutorial has been adapted from [here](https://github.com/CarloLucibello/GraphNeuralNetworks.jl/blob/master/examples/neural_ode_cora.jl).
 
-In this tutorial, we will use Graph Differential Equations (GDEs) to perform classification on the [CORA Dataset](https://relational.fit.cvut.cz/dataset/CORA). We shall be using the Graph Neural Networks primitives from the package [GraphNeuralNetworks](https://github.com/CarloLucibello/GraphNeuralNetworks.jl).
+In this tutorial, we will use Graph Differential Equations (GDEs) to perform classification on the [CORA Dataset](https://paperswithcode.com/dataset/cora). We shall be using the Graph Neural Networks primitives from the package [GraphNeuralNetworks](https://github.com/CarloLucibello/GraphNeuralNetworks.jl).
 
 ```julia
 # Load the packages

diff --git a/docs/src/examples/ode/exogenous_input.md b/docs/src/examples/ode/exogenous_input.md
@@ -40,6 +40,7 @@ In the following example, a discrete exogenous input signal `ex` is defined and
 used as an input into the neural network of a neural ODE system.
 
 ```@example exogenous
+using SciMLSensitivity
 using OrdinaryDiffEq, Lux, ComponentArrays, DiffEqFlux, Optimization,
     OptimizationPolyalgorithms, OptimizationOptimisers, Plots, Random
 
@@ -97,5 +98,3 @@ plot(tsteps, sol')
 N = length(sol)
 scatter!(tsteps, y[1:N])
 ```
-
-![](https://aws1.discourse-cdn.com/business5/uploads/julialang/original/3X/f/3/f3c2727af36ac20e114fe3c9798e567cc9d22b9e.png)
diff --git a/docs/src/examples/ode/second_order_adjoints.md b/docs/src/examples/ode/second_order_adjoints.md
@@ -14,6 +14,7 @@ with Hessian-vector products (never forming the Hessian) for large parameter
 optimizations.
 
 ```@example secondorderadjoints
+using SciMLSensitivity
 using Lux, ComponentArrays, DiffEqFlux, Optimization, OptimizationOptimisers,
     OrdinaryDiffEq, Plots, Random, OptimizationOptimJL
 

diff --git a/docs/src/examples/ode/second_order_neural.md b/docs/src/examples/ode/second_order_neural.md
@@ -21,6 +21,7 @@ neural network by the mass!)
 An example of training a neural network on a second order ODE is as follows:
 
 ```@example secondorderneural
+using SciMLSensitivity
 using OrdinaryDiffEq, Lux, Optimization, OptimizationOptimisers, RecursiveArrayTools,
     Random, ComponentArrays
 

diff --git a/docs/src/examples/optimal_control/optimal_control.md b/docs/src/examples/optimal_control/optimal_control.md
@@ -126,5 +126,3 @@ cb(res3.u, l)
 p = plot(solve(remake(prob, p = res3.u), Tsit5(), saveat = 0.01), ylim = (-6, 6), lw = 3)
 plot!(p, ts, [first(first(ann([t], ComponentArray(res3.u, ax), st))) for t in ts], label = "u(t)", lw = 3)
 ```
-
-![](https://user-images.githubusercontent.com/1814174/81859169-db65b280-9532-11ea-8394-dbb5efcd4036.png)
diff --git a/docs/src/examples/pde/pde_constrained.md b/docs/src/examples/pde/pde_constrained.md
@@ -4,6 +4,7 @@ This example uses a prediction model to optimize the one-dimensional Heat Equati
 (Step-by-step description below)
 
 ```@example pde
+using SciMLSensitivity
 using DelimitedFiles, Plots
 using OrdinaryDiffEq, Optimization, OptimizationPolyalgorithms, Zygote
 
@@ -17,7 +18,7 @@ u0 = exp.(-(x .- 3.0) .^ 2) # I.C
 
 ## Problem Parameters
 p = [1.0, 1.0]    # True solution parameters
-xtrs = [dx, Nx]      # Extra parameters
+const xtrs = [dx, Nx]      # Extra parameters
 dt = 0.40 * dx^2    # CFL condition
 t0, tMax = 0.0, 1000 * dt
 tspan = (t0, tMax)
@@ -35,22 +36,24 @@ function d2dx(u, dx)
     """
     2nd order Central difference for 2nd degree derivative
     """
-    return [[zero(eltype(u))];
-        (u[3:end] - 2.0 .* u[2:(end - 1)] + u[1:(end - 2)]) ./ (dx^2);
-        [zero(eltype(u))]]
+    return [zero(eltype(u));
+        (@view(u[3:end]) .- 2.0 .* @view(u[2:(end - 1)]) .+ @view(u[1:(end - 2)])) ./ (dx^2);
+        zero(eltype(u))]
 end
 
 ## ODE description of the Physics:
-function heat(u, p, t)
+function heat(u, p, t, xtrs)
     # Model parameters
     a0, a1 = p
     dx, Nx = xtrs #[1.0,3.0,0.125,100]
     return 2.0 * a0 .* u + a1 .* d2dx(u, dx)
 end
+heat_closure(u, p, t) = heat(u, p, t, xtrs)
 
 # Testing Solver on linear PDE
-prob = ODEProblem(heat, u0, tspan, p)
+prob = ODEProblem(heat_closure, u0, tspan, p)
 sol = solve(prob, Tsit5(), dt = dt, saveat = t);
+arr_sol = Array(sol)
 
 plot(x, sol.u[1], lw = 3, label = "t0", size = (800, 500))
 plot!(x, sol.u[end], lw = 3, ls = :dash, label = "tMax")
@@ -63,8 +66,7 @@ end
 ## Defining Loss function
 function loss(θ)
     pred = predict(θ)
-    l = predict(θ) - sol
-    return sum(abs2, l), pred # Mean squared error
+    return sum(abs2.(predict(θ) .- arr_sol)), pred # Mean squared error
 end
 
 l, pred = loss(ps)
@@ -101,6 +103,7 @@ res = Optimization.solve(optprob, PolyOpt(), callback = cb)
 ### Load Packages
 
 ```@example pde2
+using SciMLSensitivity
 using DelimitedFiles, Plots
 using OrdinaryDiffEq, Optimization, OptimizationPolyalgorithms, Zygote
 ```
@@ -122,7 +125,7 @@ u0 = exp.(-(x .- 3.0) .^ 2) # I.C
 
 ## Problem Parameters
 p = [1.0, 1.0]    # True solution parameters
-xtrs = [dx, Nx]      # Extra parameters
+const xtrs = [dx, Nx]      # Extra parameters
 dt = 0.40 * dx^2    # CFL condition
 t0, tMax = 0.0, 1000 * dt
 tspan = (t0, tMax)
@@ -159,9 +162,9 @@ function d2dx(u, dx)
     """
     2nd order Central difference for 2nd degree derivative
     """
-    return [[zero(eltype(u))];
-        (u[3:end] - 2.0 .* u[2:(end - 1)] + u[1:(end - 2)]) ./ (dx^2);
-        [zero(eltype(u))]]
+    return [zero(eltype(u));
+        (@view(u[3:end]) .- 2.0 .* @view(u[2:(end - 1)]) .+ @view(u[1:(end - 2)])) ./ (dx^2);
+        zero(eltype(u))]
 end
 ```
 
@@ -170,13 +173,13 @@ end
 Next, we set up our desired set of equations in order to define our problem.
 
 ```@example pde2
-## ODE description of the Physics:
-function heat(u, p, t)
+function heat(u, p, t, xtrs)
     # Model parameters
     a0, a1 = p
     dx, Nx = xtrs #[1.0,3.0,0.125,100]
     return 2.0 * a0 .* u + a1 .* d2dx(u, dx)
 end
+heat_closure(u, p, t) = heat(u, p, t, xtrs)
 ```
 
 ### Solve and Plot Ground Truth
@@ -186,8 +189,9 @@ will compare to further on.
 
 ```@example pde2
 # Testing Solver on linear PDE
-prob = ODEProblem(heat, u0, tspan, p)
+prob = ODEProblem(heat_closure, u0, tspan, p)
 sol = solve(prob, Tsit5(), dt = dt, saveat = t);
+arr_sol = Array(sol)
 
 plot(x, sol.u[1], lw = 3, label = "t0", size = (800, 500))
 plot!(x, sol.u[end], lw = 3, ls = :dash, label = "tMax")
@@ -222,8 +226,7 @@ use the **mean squared error**.
 ## Defining Loss function
 function loss(θ)
     pred = predict(θ)
-    l = predict(θ) - sol
-    return sum(abs2, l), pred # Mean squared error
+    return sum(abs2.(predict(θ) .- arr_sol)), pred # Mean squared error
 end
 
 l, pred = loss(ps)

diff --git a/docs/src/tutorials/data_parallel.md b/docs/src/tutorials/data_parallel.md
@@ -86,7 +86,7 @@ The following is a full copy-paste example for the multithreading.
 Distributed and GPU minibatching are described below.
 
 ```@example dataparallel
-using OrdinaryDiffEq, Optimization, OptimizationOptimisers
+using OrdinaryDiffEq, Optimization, OptimizationOptimisers, SciMLSensitivity
 pa = [1.0]
 u0 = [3.0]
 θ = [u0; pa]

diff --git a/docs/src/tutorials/training_tips/divergence.md b/docs/src/tutorials/training_tips/divergence.md
@@ -28,6 +28,7 @@ end
 A full example making use of this trick is:
 
 ```@example divergence
+using SciMLSensitivity
 using OrdinaryDiffEq, SciMLSensitivity, Optimization, OptimizationOptimisers,
     OptimizationNLopt, Plots
 

diff --git a/docs/src/tutorials/training_tips/local_minima.md b/docs/src/tutorials/training_tips/local_minima.md
@@ -16,6 +16,7 @@ before, except with one small twist: we wish to find the neural ODE that fits
 on `(0,5.0)`. Naively, we use the same training strategy as before:
 
 ```@example iterativefit
+using SciMLSensitivity
 using OrdinaryDiffEq,
     ComponentArrays, SciMLSensitivity, Optimization, OptimizationOptimisers
 using Lux, Plots, Random, Zygote
@@ -166,6 +167,7 @@ time span and (0, 10), any longer and more iterations will be required. Alternat
 one could use a mix of (3) and (4), or breaking up the trajectory into chunks and just (4).
 
 ```@example resetic
+using SciMLSensitivity
 using OrdinaryDiffEq,
     ComponentArrays, SciMLSensitivity, Optimization, OptimizationOptimisers
 using Lux, Plots, Random, Zygote

diff --git a/docs/src/tutorials/training_tips/multiple_nn.md b/docs/src/tutorials/training_tips/multiple_nn.md
@@ -7,6 +7,7 @@ this kind of study.
 The following is a fully working demo on the Fitzhugh-Nagumo ODE:
 
 ```@example
+using SciMLSensitivity
 using Lux, DiffEqFlux, ComponentArrays, Optimization, OptimizationNLopt,
     OptimizationOptimisers, OrdinaryDiffEq, Random