Add random temporal hyperbolic graph

src/GNNGraphs/GNNGraphs.jl

@@ -90,7 +90,8 @@ export rand_graph,
        rand_bipartite_heterograph,
        knn_graph,
        radius_graph,
-       rand_temporal_radius_graph
+       rand_temporal_radius_graph,
+       rand_temporal_hyperbolic_graph
 
 include("sampling.jl")
 export sample_neighbors

src/GNNGraphs/generate.jl

@@ -362,3 +362,99 @@ function rand_temporal_radius_graph(number_nodes::Int,
     end
     return TemporalSnapshotsGNNGraph(tg)
 end
+
+
+function _hyperbolic_distance(nodeA::Array{Float64, 1},nodeB::Array{Float64, 1}; ζ::Real)
+    if nodeA != nodeB
+        a = cosh(ζ * nodeA[1]) * cosh(ζ * nodeB[1])
+        b = sinh(ζ * nodeA[1]) * sinh(ζ * nodeB[1])
+        c = cos(pi - abs(pi - abs(nodeA[2] - nodeB[2])))
+        d = acosh(a - (b * c)) / ζ
+    else
+        d = 0.0
+    end
+    return d
+end
+
+"""
+    rand_temporal_hyperbolic_graph(number_nodes::Int, 
+                                   number_snapshots::Int;
+                                   α::Real,
+                                   R::Real,
+                                   speed::Real,
+                                   ζ::Real=1,
+                                   self_loop = false,
+                                   kws...)
+
+Create a random temporal graph given `number_nodes` nodes and `number_snapshots` snapshots.
+First, the positions of the nodes are generated with a quasi-uniform distribution (depending on the parameter `α`) in hyperbolic space within a disk of radius `R`. Two nodes are connected if their hyperbolic distance is less than `R`. Each following snapshot is created in order to keep the same initial distribution.
+
+# Arguments
+
+- `number_nodes`: The number of nodes of each snapshot.
+- `number_snapshots`: The number of snapshots.
+- `α`: The parameter that controls the position of the points. If `α=ζ`, the points are uniformly distributed on the disk of radius `R`. If `α>ζ`, the points are more concentrated in the center of the disk. If `α<ζ`, the points are more concentrated at the boundary of the disk.
+- `R`: The radius of the disk and of connection.
+- `speed`: The speed to update the nodes.
+- `ζ`: The parameter that controls the curvature of the disk.
+- `self_loops`: If `true`, consider the node itself among its neighbors, in which
+                case the graph will contain self-loops.
+- `kws`: Further keyword arguments will be passed to the [`GNNGraph`](@ref) constructor of each snapshot.
+
+# Example
+
+```julia-repl
+julia> n, snaps, α, R, speed, ζ = 10, 5, 1.0, 4.0, 0.1, 1.0;
+
+julia> thg = rand_temporal_hyperbolic_graph(n, snaps; α, R, speed, ζ)
+TemporalSnapshotsGNNGraph:
+  num_nodes: [10, 10, 10, 10, 10]
+  num_edges: [44, 46, 48, 42, 38]
+  num_snapshots: 5
+```
+
+# References
+Section D of the paper [Dynamic Hidden-Variable Network Models](https://arxiv.org/pdf/2101.00414.pdf) and the paper 
+[Hyperbolic Geometry of Complex Networks](https://arxiv.org/pdf/1006.5169.pdf)
+"""
+function rand_temporal_hyperbolic_graph(number_nodes::Int,
+                                        number_snapshots::Int;
+                                        α::Real,
+                                        R::Real,
+                                        speed::Real,
+                                        ζ::Real=1,
+                                        self_loop = false,
+                                        kws...)
+        @assert number_snapshots > 1 "The number of snapshots must be greater than 1"
+        @assert α > 0 "α must be greater than 0"
+
+        probabilities = rand(number_nodes)
+
+        points = Array{Float64}(undef,2,number_nodes)
+        points[1,:].= (1/α) * acosh.(1 .+ (cosh(α * R) - 1) * probabilities)
+        points[2,:].= 2 * pi * rand(number_nodes)
+
+        tg = Vector{GNNGraph}(undef, number_snapshots)
+
+        for time in 1:number_snapshots
+            adj = zeros(number_nodes,number_nodes)
+            for i in 1:number_nodes
+                for j in 1:number_nodes
+                    if !self_loop && i==j
+                        continue
+                    elseif _hyperbolic_distance(points[:,i],points[:,j]; ζ) <= R
+                        adj[i,j] = adj[j,i] = 1
+                    end
+                end
+            end
+            tg[time] = GNNGraph(adj)
+
+            probabilities .= probabilities .+ (2 * speed * rand(number_nodes) .- speed)
+            probabilities[probabilities.>1] .=  1 .- (probabilities[probabilities .> 1] .% 1)
+            probabilities[probabilities.<0] .= abs.(probabilities[probabilities .< 0])
+
+            points[1,:].= (1/α) * acosh.(1 .+ (cosh(α * R) - 1) * probabilities)
+            points[2,:].= points[2,:] .+ (2 * speed * rand(number_nodes) .- speed)
+        end
+    return TemporalSnapshotsGNNGraph(tg)
+end

test/GNNGraphs/generate.jl

@@ -104,4 +104,19 @@ end
     r2 = 0.95
     tg2 = rand_temporal_radius_graph(number_nodes, number_snapshots, speed, r2)
     @test mean(mean(degree.(tg.snapshots)))<=mean(mean(degree.(tg2.snapshots)))
+end
+
+@testset "rand_temporal_hyperbolic_graph" begin
+    @test GraphNeuralNetworks.GNNGraphs._hyperbolic_distance([1.0,1.0],[1.0,1.0];ζ=1)==0
+    @test GraphNeuralNetworks.GNNGraphs._hyperbolic_distance([0.23,0.11],[0.98,0.55];ζ=1)==GraphNeuralNetworks.GNNGraphs._hyperbolic_distance([0.98,0.55],[0.23,0.11];ζ=1)
+    number_nodes = 30
+    number_snapshots = 5
+    α, R, speed, ζ = 1, 1, 0.1, 1
+
+    tg = rand_temporal_hyperbolic_graph(number_nodes, number_snapshots; α, R, speed, ζ)
+    @test tg.num_nodes == [number_nodes for i in 1:number_snapshots]
+    @test tg.num_snapshots == number_snapshots
+    R = 10
+    tg1 = rand_temporal_hyperbolic_graph(number_nodes, number_snapshots; α, R, speed, ζ)
+    @test mean(mean(degree.(tg1.snapshots)))<=mean(mean(degree.(tg.snapshots)))
 end