Finish the separation from GeometricOptimizers

Project.toml

@@ -1,6 +1,6 @@
 name = "GeometricMachineLearning"
 uuid = "194d25b2-d3f5-49f0-af24-c124f4aa80cc"
-version = "0.5.0-DEV"
+version = "0.5.0"
 authors = ["Michael Kraus <michael.kraus@ipp.mpg.de>"]
 
 [deps]
@@ -41,7 +41,7 @@ ForwardDiff = "0.10, 1"
 GeometricBase = "0.14"
 GeometricEquations = "0.21"
 GeometricIntegrators = "0.18.2"
-GeometricOptimizers = "0.3"
+GeometricOptimizers = "0.4"
 GeometricSolutions = "0.6"
 HDF5 = "0.16, 0.17"
 KernelAbstractions = "0.9"

README.md

@@ -18,6 +18,8 @@ At its core every neural network comprises three components: a neural network ar
 
 Traditionally, physical properties have been encoded into the loss function (PINN approach), but in `GeometricMachineLearning.jl` this is exclusively done through the architectures and the optimizers of the neural network, thus giving theoretical guarantees that these properties are actually preserved.
 
+The optimizer methods themselves — `GradientMethod`, `MomentumMethod`, `Adam`, the manifold types they act on, and the caches, global sections and retractions that go with them — come from [`GeometricOptimizers.jl`](https://github.com/JuliaGNI/GeometricOptimizers.jl) and are re-exported here. `GeometricMachineLearning.jl` supplies the part that is specific to neural networks: the architectures, the layers, and walking the parameter tree of a network during training.
+
 Using the package is very straightforward and is very flexible with respect to the device `(CPU, CUDA, Metal, ...)` and the type `(Float16, Float32, Float64, ...)` you want to use. The following is a simple example to learn a SympNet on data coming from a pendulum:
 ```julia
 using GeometricMachineLearning
@@ -41,8 +43,8 @@ backend = CUDABackend()
 # initialize the network (i.e. the parameters of the network)
 g_nn = NeuralNetwork(gsympnet, backend, type)
 
-# call the optimizer
-g_opt = Optimizer(AdamOptimizer(), g_nn)
+# call the optimizer: the method comes first, the step size is given separately
+g_opt = Optimizer(Adam(type), g_nn; step_size = 1e-3)
 
 const nepochs = 300
 const batch_size = 100
@@ -53,7 +55,7 @@ g_loss_array = g_opt(g_nn, dl, Batch(batch_size), nepochs)
 # plot the result
 ics = (q=qp_data.q[:,1], p=qp_data.p[:,1])
 const steps_to_plot = 200
-g_trajectory = Iterate_Sympnet(g_nn, ics; n_points = steps_to_plot)
+g_trajectory = iterate(g_nn, ics; n_points = steps_to_plot)
 fig = Figure()
 ax = Axis(fig[1, 1]; xlabel = "q", ylabel = "p")
 lines!(ax, vec(qp_data.q')[1:steps_to_plot], vec(qp_data.p')[1:steps_to_plot]; label = "training data")

docs/Makefile

@@ -121,11 +121,15 @@ remove_numbering_for_some_chapters:
 	sed -i'' -e 's/\\section{Chapter Summary}/\\section*{Chapter Summary}/g' build/G*.tex;
 	sed -i'' -e 's/\\section{References}/\\section*{References}/g' build/G*.tex;
 
+# `build/manifolds` and `build/optimizers/manifold_related` used to be listed here. Documenter only
+# creates a `build/` subdirectory for a page tree that still has pages, so those two went away with
+# the chapters that moved to GeometricOptimizers -- and `find` exits non-zero on a missing root,
+# which aborts the recipe and takes the PDF workflow with it. The roots that remain are guarded for
+# the same reason, so that the next chapter to move out does not break the build again.
 copy_png_files:
-	find build/manifolds -name \*.png -exec cp {} build \; ;
-	find build/optimizers/manifold_related -name \*.png -exec cp {} build \; ;
-	find build/tutorials -name \*.png -exec cp {} build \;
-	find build/reduced_order_modeling -name \*.png -exec cp {} build \;
+	for d in tutorials reduced_order_modeling; do \
+		if [ -d "build/$$d" ]; then find "build/$$d" -name \*.png -exec cp {} build \; ; fi; \
+	done
 
 docstring_indexing:
 	sed -i'' -e 's/\\item \\hyperlinkref{\([0-9]*\)}{\\texttt{\(.*\)}}/\\item \\hyperlinkref{\1}{\\texttt{\2}}: page \\pageref{\1}/g' build/G*.tex;

docs/Project.toml

@@ -3,8 +3,10 @@ AbstractNeuralNetworks = "60874f82-5ada-4c70-bd1c-fa6be7711c8a"
 Bibliography = "f1be7e48-bf82-45af-a471-ae754a193061"
 BrenierTwoFluid = "698bc5df-bacc-4e45-9592-41ae9e406d75"
 CairoMakie = "13f3f980-e62b-5c42-98c6-ff1f3baf88f0"
+DocInventories = "43dc2714-ed3b-44b5-b226-857eda1aa7de"
 Documenter = "e30172f5-a6a5-5a46-863b-614d45cd2de4"
 DocumenterCitations = "daee34ce-89f3-4625-b898-19384cb65244"
+DocumenterInterLinks = "d12716ef-a0f6-4df4-a9f1-a5a34e75c656"
 ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
 GeometricIntegrators = "dcce2d33-59f6-5b8d-9047-0defad88ae06"
 GeometricMachineLearning = "194d25b2-d3f5-49f0-af24-c124f4aa80cc"
@@ -14,8 +16,6 @@ HDF5 = "f67ccb44-e63f-5c2f-98bd-6dc0ccc4ba2f"
 LaTeXStrings = "b964fa9f-0449-5b57-a5c2-d3ea65f4040f"
 Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 
-# GeometricMachineLearning itself is deliberately *not* listed here: `docs/Makefile`'s `test_docs`
-# target develops it from the repository root, which is what makes the same file work in CI and in a
-# local clone at any path.
 [sources]
 BrenierTwoFluid = {rev = "main", url = "https://github.com/ToBlick/BrenierTwoFluids.git"}
+GeometricMachineLearning = {path = ".."}