From c779fc589d5e86409ae7dec344e927280fbc2eed Mon Sep 17 00:00:00 2001
From: abieler <andre.bieler81@gmail.com>
Date: Thu, 15 Jul 2021 14:51:22 +0200
Subject: [PATCH] Add more example scripts

---
 examples/gcn.jl                | 17 ++++++------
 examples/gcn_featured_graph.jl | 50 ++++++++++++++++++++++++++++++++++
 examples/gcn_gpu.jl            | 47 ++++++++++++++++++++++++++++++++
 3 files changed, 106 insertions(+), 8 deletions(-)
 create mode 100644 examples/gcn_featured_graph.jl
 create mode 100644 examples/gcn_gpu.jl

diff --git a/examples/gcn.jl b/examples/gcn.jl
index ecf641558..acd4399bd 100644
--- a/examples/gcn.jl
+++ b/examples/gcn.jl
@@ -23,15 +23,16 @@ target_catg = 7
 epochs = 100
 
 ## Preprocessing data
-train_X = Matrix{Float32}(features) |> gpu  # dim: num_features * num_nodes
-train_y = Matrix{Float32}(labels) |> gpu  # dim: target_catg * num_nodes
-adj_mat = Matrix{Float32}(adjacency_matrix(g)) |> gpu
-
+train_X = Matrix{Float32}(features)  # dim: num_features * num_nodes
+train_y = Matrix{Float32}(labels) # dim: target_catg * num_nodes
+adj_mat = Matrix{Float32}(adjacency_matrix(g))
+
+model = Chain(
+    GCNConv(adj_mat, num_features=>hidden, relu),
+    Dropout(0.5),
+    GCNConv(adj_mat, hidden=>target_catg),
+)
 ## Model
-model = Chain(GCNConv(adj_mat, num_features=>hidden, relu),
-              Dropout(0.5),
-              GCNConv(adj_mat, hidden=>target_catg),
-              ) |> gpu
 
 ## Loss
 loss(x, y) = logitcrossentropy(model(x), y)
diff --git a/examples/gcn_featured_graph.jl b/examples/gcn_featured_graph.jl
new file mode 100644
index 000000000..4c87f7b80
--- /dev/null
+++ b/examples/gcn_featured_graph.jl
@@ -0,0 +1,50 @@
+using GeometricFlux
+using Flux
+using Flux: onehotbatch, onecold, logitcrossentropy, throttle
+using Flux: @epochs
+using JLD2
+using Statistics
+using SparseArrays
+using LightGraphs.SimpleGraphs
+using LightGraphs: adjacency_matrix
+using CUDA
+using Random
+
+Random.seed!([0x6044b4da, 0xd873e4f9, 0x59d90c0a, 0xde01aa81])
+
+@load "data/cora_features.jld2" features
+@load "data/cora_labels.jld2" labels
+@load "data/cora_graph.jld2" g
+
+num_nodes = 2708
+num_features = 1433
+hidden = 16
+target_catg = 7
+epochs = 5
+
+## Preprocessing data
+train_X = Matrix{Float32}(features)  # dim: num_features * num_nodes
+train_y = Matrix{Float32}(labels) # dim: target_catg * num_nodes
+adj_mat = Matrix{Float32}(adjacency_matrix(g))
+
+## Model
+model = Chain(
+    GCNConv(num_features=>hidden, relu),
+    # Dropout(0.5),  --> does not work
+    GCNConv(hidden=>target_catg, relu),
+    FeatureSelector(:node)
+)
+
+## Loss
+loss(x, y) = logitcrossentropy(model(x), y)
+accuracy(x, y) = mean(onecold(softmax(cpu(model(x)))) .== onecold(cpu(y)))
+
+
+## Training
+ps = Flux.params(model)
+fg = FeaturedGraph(adj_mat, nf=train_X)
+train_data = [(fg, train_y)]
+opt = ADAM(0.01)
+evalcb() = @show(accuracy(fg, train_y))
+
+@epochs epochs Flux.train!(loss, ps, train_data, opt, cb=throttle(evalcb, 10))
diff --git a/examples/gcn_gpu.jl b/examples/gcn_gpu.jl
new file mode 100644
index 000000000..9a95dd5f5
--- /dev/null
+++ b/examples/gcn_gpu.jl
@@ -0,0 +1,47 @@
+using GeometricFlux
+using Flux
+using Flux: onehotbatch, onecold, logitcrossentropy, throttle
+using Flux: @epochs
+using JLD2
+using Statistics
+using SparseArrays
+using LightGraphs.SimpleGraphs
+using LightGraphs: adjacency_matrix
+using CUDA
+using Random
+
+Random.seed!([0x6044b4da, 0xd873e4f9, 0x59d90c0a, 0xde01aa81])
+
+@load "data/cora_features.jld2" features
+@load "data/cora_labels.jld2" labels
+@load "data/cora_graph.jld2" g
+
+num_nodes = 2708
+num_features = 1433
+hidden = 16
+target_catg = 7
+epochs = 100
+
+## Preprocessing data
+train_X = Matrix{Float32}(features) |> gpu             # dim: num_features * num_nodes
+train_y = Matrix{Float32}(labels) |> gpu               # dim: target_catg * num_nodes
+adj_mat = Matrix{Float32}(adjacency_matrix(g)) |> gpu
+
+## Model
+model = Chain(GCNConv(adj_mat, num_features=>hidden, relu),
+              Dropout(0.5),
+              GCNConv(adj_mat, hidden=>target_catg),
+              ) |> gpu
+
+## Loss
+loss(x, y) = logitcrossentropy(model(x), y)
+accuracy(x, y) = mean(onecold(softmax(cpu(model(x)))) .== onecold(cpu(y)))
+
+
+## Training
+ps = Flux.params(model)
+train_data = [(train_X, train_y)]
+opt = ADAM(0.01)
+evalcb() = @show(accuracy(train_X, train_y))
+
+@epochs epochs Flux.train!(loss, ps, train_data, opt, cb=throttle(evalcb, 10))