add docs

CarloLucibello · CarloLucibello · commit 682b1b7808d7 · 2021-08-31T09:26:11.000+02:00
diff --git a/README.md b/README.md
@@ -24,4 +24,3 @@ Some of its noticeable features are the following:
 ## Usage
 
 Usage examples can be found in the `examples/` folder. 
-```
diff --git a/docs/make.jl b/docs/make.jl
@@ -6,10 +6,18 @@ DocMeta.setdocmeta!(GraphNeuralNetworks, :DocTestSetup, :(using GraphNeuralNetwo
 makedocs(;
     modules=[GraphNeuralNetworks],
     sitename = "GraphNeuralNetworks.jl",
-    pages=[
-        "Home" => "index.md",
-        "Developer Notes" => "dev.md",
-    ],
+    pages = ["Home" => "index.md",
+             "Graphs" => "graphs.md",
+             "Message passing" => "messagepassing.md",
+             "Building models" => "models.md",
+             "API Reference" =>
+               [
+                "Graphs" => "api/graphs.md",
+                "Convolutional Layers" => "api/conv.md",
+                "Pooling Layers" => "api/pool.md",
+               ],
+              "Developer Notes" => "dev.md",
+            ],
 )
 
 deploydocs(repo="github.com/CarloLucibello/GraphNeuralNetworks.jl.git")
diff --git a/docs/src/api/conv.md b/docs/src/api/conv.md
@@ -0,0 +1,12 @@
+# Convolutional Layers
+
+```@index
+Order = [:type, :function]
+Pages   = ["api/conv.md"]
+```
+
+```@autodocs
+Modules = [GraphNeuralNetworks]
+Pages   = ["layers/conv.jl"]
+Private = false
+```
diff --git a/docs/src/api/graphs.md b/docs/src/api/graphs.md
@@ -0,0 +1,12 @@
+# Graphs
+
+```@index
+Order = [:type, :function]
+Pages   = ["api/graphs.md"]
+```
+
+```@autodocs
+Modules = [GraphNeuralNetworks]
+Pages   = ["featuredgraph.jl"]
+Private = false
+```
diff --git a/docs/src/api/pool.md b/docs/src/api/pool.md
@@ -0,0 +1,12 @@
+# Pooling Layers
+
+```@index
+Order = [:type, :function]
+Pages   = ["api/pool.md"]
+```
+
+```@autodocs
+Modules = [GraphNeuralNetworks]
+Pages   = ["layers/pool.jl"]
+Private = false
+```
diff --git a/docs/src/graphs.md b/docs/src/graphs.md
@@ -0,0 +1,2 @@
+# Graphs
+
diff --git a/docs/src/index.md b/docs/src/index.md
@@ -1,14 +1,7 @@
-```@meta
-CurrentModule = GraphNeuralNetworks
-```
-
 # GraphNeuralNetworks
 
 Documentation for [GraphNeuralNetworks](https://github.com/CarloLucibello/GraphNeuralNetworks.jl).
 
-```@index
-```
 
-```@autodocs
-Modules = [GraphNeuralNetworks]
-```
+## Getting Started
+
diff --git a/docs/src/messagepassing.md b/docs/src/messagepassing.md
@@ -0,0 +1,2 @@
+# Message Passing
+
diff --git a/docs/src/models.md b/docs/src/models.md
@@ -0,0 +1 @@
+# Models
diff --git a/src/featuredgraph.jl b/src/featuredgraph.jl
@@ -26,7 +26,7 @@ is preserved and shared.
 A `FeaturedGraph` is a LightGraphs' `AbstractGraph`, therefore any functionality
 from the LightGraphs' graph library can be used on it.
 
-# Arguments 
+## Arguments 
 
 - `g`: Some data representing the graph topology. Possible type are 
     - An adjacency matrix
@@ -302,7 +302,7 @@ end
 
 Normalized Laplacian matrix of graph `g`.
 
-# Arguments
+## Arguments
 
 - `fg`: A `FeaturedGraph`.
 - `T`: result element type.
@@ -332,7 +332,7 @@ end
 Scaled Laplacian matrix of graph `g`,
 defined as ``\hat{L} = \frac{2}{\lambda_{max}} L - I`` where ``L`` is the normalized Laplacian matrix.
 
-# Arguments
+## Arguments
 
 - `fg`: A `FeaturedGraph`.
 - `T`: result element type.
diff --git a/src/layers/conv.jl b/src/layers/conv.jl
@@ -1,18 +1,24 @@
-"""
+@doc raw"""
     GCNConv(in => out, σ=identity; bias=true, init=glorot_uniform)
 
-Graph convolutional layer.
+Graph convolutional layer from paper [Semi-supervised Classification with Graph Convolutional Networks](https://arxiv.org/abs/1609.02907).
 
-# Arguments
+Performs the operation
+```math
+\mathbf{x}'_i = \sum_{j\in N(i)} \frac{1}{c_{ij}} W \mathbf{x}_j
+```
+where ``c_{ij} = \sqrt{N(i)\,N(j)}``.
 
-- `in`: The dimension of input features.
-- `out`: The dimension of output features.
+The input to the layer is a node feature array `X` 
+of size `(num_features, num_nodes)`.
+
+## Arguments
+
+- `in`: Number of input features.
+- `out`: Number of output features.
 - `σ`: Activation function.
 - `bias`: Add learnable bias.
 - `init`: Weights' initializer.
-
-The input to the layer is a node feature array `X` 
-of size `(num_features, num_nodes)`.
 """
 struct GCNConv{A<:AbstractMatrix, B, F} <: MessagePassing
     weight::A
@@ -61,12 +67,29 @@ function Base.show(io::IO, l::GCNConv)
 end
 
 
-"""
+@doc raw"""
     ChebConv(in=>out, k; bias=true, init=glorot_uniform)
 
-Chebyshev spectral graph convolutional layer.
+Chebyshev spectral graph convolutional layer from
+paper [Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering](https://arxiv.org/abs/1606.09375).
+
+Implements
+
+```math
+X' = \sum^{K-1}_{k=0}  W^{(k)} Z^{(k)}
+```
+
+where ``Z^{(k)}`` is the ``k``-th term of Chebyshev polynomials, and can be calculated by the following recursive form:
+
+```math
+Z^{(0)} = X \\
+Z^{(1)} = \hat{L} X \\
+Z^{(k)} = 2 \hat{L} Z^{(k-1)} - Z^{(k-2)}
+```
 
-# Arguments
+with ``\hat{L}`` the [`scaled_laplacian`](@ref).
+
+## Arguments
 
 - `in`: The dimension of input features.
 - `out`: The dimension of output features.
@@ -117,18 +140,23 @@ function Base.show(io::IO, l::ChebConv)
 end
 
 
-"""
+@doc raw"""
     GraphConv(in => out, σ=identity, aggr=+; bias=true, init=glorot_uniform)
 
-Graph neural network layer.
+Graph convolution layer from Reference: [Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networks](https://arxiv.org/abs/1810.02244).
 
-# Arguments
+Performs:
+```math
+\mathbf{x}_i' = W^1 \mathbf{x}_i + \box_{j \in \mathcal{N}(i)} W^2 \mathbf{x}_j)
+```
+where the aggregation type is selected by `aggr`.
+
+## Arguments
 
 - `in`: The dimension of input features.
 - `out`: The dimension of output features.
 - `σ`: Activation function.
-- `aggr`: An aggregate function applied to the result of message function. `+`, `-`,
-`*`, `/`, `max`, `min` and `mean` are available.
+- `aggr`: Aggregation operator for the incoming messages (e.g. `+`, `*`, `max`, `min`, and `mean`).
 - `bias`: Add learnable bias.
 - `init`: Weights' initializer.
 """
@@ -172,17 +200,27 @@ function Base.show(io::IO, l::GraphConv)
 end
 
 
-"""
+@doc raw"""
     GATConv(in => out;
             heads=1,
             concat=true,
             init=glorot_uniform    
             bias=true, 
             negative_slope=0.2)
 
-Graph attentional layer.
+Graph attentional layer from the paper [Graph Attention Networks](https://arxiv.org/abs/1710.10903).
 
-# Arguments
+Implements the operation
+```math
+\mathbf{x}_i' = \sum_{j \in N(i)} \alpha_{ij} W \mathbf{x}_j
+```
+where the attention coefficient ``\alpha_{ij}`` is given by
+```math
+\alpha_{ij} = \frac{1}{z_i} exp(LeakyReLU(\mathbf{a}^T [W \mathbf{x}_i || W \mathbf{x}_j]))
+```
+with ``z_i`` a normalization factor.
+
+## Arguments
 
 - `in`: The dimension of input features.
 - `out`: The dimension of output features.
@@ -255,17 +293,25 @@ function Base.show(io::IO, l::GATConv)
 end
 
 
-"""
+@doc raw"""
     GatedGraphConv(out, num_layers; aggr=+, init=glorot_uniform)
 
-Gated graph convolution layer.
+Gated graph convolution layer from [Gated Graph Sequence Neural Networks](https://arxiv.org/abs/1511.05493).
+
+Implements the recursion
+```math
+\mathbf{h}^{(0)}_i = \mathbf{x}_i || \mathbf{0} \\
+\mathbf{h}^{(l)}_i = GRU(\mathbf{h}^{(l-1)}_i, \box_{j \in N(i)} W \mathbf{h}^{(l-1)}_j)
+```
 
-# Arguments
+ where ``\mathbf{h}^{(l)}_i`` denotes the ``l``-th hidden variables passing through GRU. The dimension of input ``\mathbf{x}_i`` needs to be less or equal to `out`.
+
+## Arguments
 
 - `out`: The dimension of output features.
 - `num_layers`: The number of gated recurrent unit.
-- `aggr`: An aggregate function applied to the result of message function. `+`, `-`,
-`*`, `/`, `max`, `min` and `mean` are available.
+- `aggr`: Aggregation operator for the incoming messages (e.g. `+`, `*`, `max`, `min`, and `mean`).
+- `init`: Weight initialization function.
 """
 struct GatedGraphConv{A<:AbstractArray{<:Number,3}, R} <: MessagePassing
     weight::A
@@ -313,16 +359,22 @@ function Base.show(io::IO, l::GatedGraphConv)
 end
 
 
+@doc raw"""
+    EdgeConv(f; aggr=max)
 
-"""
-    EdgeConv(nn; aggr=max)
+Edge convolutional layer from paper [Dynamic Graph CNN for Learning on Point Clouds](https://arxiv.org/abs/1801.07829).
 
-Edge convolutional layer.
+Performs the operation
+```math
+\mathbf{x}_i' = \box_{j \in N(i)} f(\mathbf{x}_i || \mathbf{x}_j - \mathbf{x}_i)
+```
 
-# Arguments
+where `f` typically denotes a learnable function, e.g. a linear layer or a multi-layer perceptron.
 
-- `nn`: A neural network (e.g. a Dense layer or a MLP). 
-- `aggr`: An aggregate function applied to the result of message function. `+`, `max` and `mean` are available.
+## Arguments
+
+- `f`: A (possibly learnable) function acting on edge features. 
+- `aggr`: Aggregation operator for the incoming messages (e.g. `+`, `*`, `max`, `min`, and `mean`).
 """
 struct EdgeConv <: MessagePassing
     nn
@@ -352,18 +404,21 @@ function Base.show(io::IO, l::EdgeConv)
 end
 
 
-"""
-    GINConv(nn; eps = 0f0)
+@doc raw"""
+    GINConv(f; eps = 0f0)
 
-Graph Isomorphism Network.
+Graph Isomorphism convolutional layer from paper [How Powerful are Graph Neural Networks?](https://arxiv.org/pdf/1810.00826.pdf)
 
-# Arguments
 
-- `nn`: A neural network/layer.
-- `eps`: Weighting factor.
+```math
+\mathbf{x}_i' = f\left((1 + \epsilon) \mathbf{x}_i + \sum_{j \in N(i)} \mathbf{x}_j \right)
+```
+where `f` typically denotes a learnable function, e.g. a linear layer or a multi-layer perceptron.
+
+## Arguments
 
-The definition of this is as defined in the original paper,
-Xu et. al. (2018) https://arxiv.org/abs/1810.00826.
+- `f`: A (possibly learnable) function acting on node features. 
+- `eps`: Weighting factor.
 """
 struct GINConv{R<:Real} <: MessagePassing
     nn
diff --git a/src/layers/msgpass.jl b/src/layers/msgpass.jl
@@ -74,7 +74,7 @@ By default, the function returns `x_j`.
 Layers subtyping [`MessagePassing`](@ref) should 
 specialize this method with custom behavior.
 
-# Arguments
+## Arguments
 
 - `mp`: A [`MessagePassing`](@ref) layer.
 - `x_i`: Features of the central node `i`.
@@ -98,7 +98,7 @@ By default, the function returns `m̄`.
 Layers subtyping [`MessagePassing`](@ref) should 
 specialize this method with custom behavior.
 
-# Arguments
+## Arguments
 
 - `mp`: A [`MessagePassing`](@ref) layer.
 - `m̄`: Aggregated edge messages from the [`message`](@ref) function.
diff --git a/src/layers/pool.jl b/src/layers/pool.jl
@@ -7,7 +7,7 @@ Global pooling layer.
 
 It pools all features with `aggr` operation.
 
-# Arguments
+## Arguments
 
 - `aggr`: An aggregate function applied to pool all features.
 """
@@ -29,7 +29,7 @@ Local pooling layer.
 
 It pools features with `aggr` operation accroding to `cluster`. It is implemented with `scatter` operation.
 
-# Arguments
+## Arguments
 
 - `aggr`: An aggregate function applied to pool all features.
 - `cluster`: An index structure which indicates what features to aggregate with.
@@ -46,7 +46,7 @@ end
 
 Top-k pooling layer.
 
-# Arguments
+## Arguments
 
 - `adj`: Adjacency matrix  of a graph.
 - `k`: Top-k nodes are selected to pool together.

Original file line number	Diff line number	Diff line change
`@@ -24,4 +24,3 @@ Some of its noticeable features are the following:`
`24`	`24`	`## Usage`
`25`	`25`
`26`	`26`	Usage examples can be found in the `examples/` folder.
`27`		-```