-
Notifications
You must be signed in to change notification settings - Fork 3
/
run_dataset.jl
177 lines (151 loc) · 6.95 KB
/
run_dataset.jl
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
using Random;
using Statistics;
using StatsBase: sample, randperm, mean;
using LinearAlgebra;
using SparseArrays;
using IterativeSolvers;
using LightGraphs;
using Flux;
using GraphSAGE;
using BSON: @save, @load;
using Printf;
include("read_network.jl");
include("utils.jl");
function run_dataset(network_trans, network_ind, accuracyFun, regressor="zero", correlation="zero", inductive=false)
@assert regressor in ["zero", "linear", "mlp", "gnn"]
@assert correlation in ["zero", "homo", "learned"]
dim_out, dim_h = 8, 16;
t, k, num_steps = 128, 32, 1500;
num_ave = 10;
ptr = 0.60;
rrt = Vector();
αt = Vector();
βt = Vector();
if inductive
ptr_inductive = 0.00:0.05:0.60;
rri = [[Vector() for _ in 1:length(ptr_inductive)] for _ in 1:length(network_ind)];
end
function run_once(seed_val)
Random.seed!(seed_val);
println("\n\nseed_val: ", seed_val);
G, A, labels, feats = read_network(network_trans); n = nv(G);
d = sum(sum(A), dims=1)[:];
S = [spdiagm(0=>d.^-0.5)*A_*spdiagm(0=>d.^-0.5) for A_ in A];
L, VU = rand_split(n, ptr);
V, U = VU[1:div(length(VU),2)], VU[div(length(VU),2)+1:end];
ab = param(vcat(zeros(length(A)), 3.0));
getα() = tanh.(ab[1:end-1]);
getβ() = exp(ab[end]);
if regressor == "zero"
getRegression = L -> zeros(length(L));
θ = params();
optθ = ADAM(0.0);
elseif regressor == "linear"
lls = Dense(length(feats[1]), 1);
getRegression = L -> vcat(lls.([feats[u] for u in L])...);
θ = params(lls);
optθ = ADAM(0.1);
elseif regressor == "mlp"
mlp = Chain(Dense(length(feats[1]), dim_h, relu), Dense(dim_h, dim_h, relu), Dense(dim_h, dim_out, relu), Dense(dim_out, 1));
getRegression = L -> vcat(mlp.([feats[u] for u in L])...);
θ = params(mlp);
optθ = ADAM(0.001);
elseif regressor == "gnn"
enc = graph_encoder(length(feats[1]), dim_out, dim_h, repeat(["SAGE_Mean"], 2); σ=relu);
reg = Dense(dim_out, 1);
getRegression = L -> vcat(reg.(enc(G, L, u->feats[u]))...);
θ = params(enc, reg);
optθ = ADAM(0.001);
else
error("unexpected regressor type");
end
if match(r"twitch", network_trans) != nothing
optφ = "l_bfgs";
φ_skip = 100;
else
optφ = Descent(0.1);
φ_skip = 10;
end
getrL(L) = labels[L] - getRegression(L);
function getΩ(α, β, rL, L, logdet)
Ω = quadformSC(α, β, rL; S=S, L=L);
logdet && (Ω -= (logdetΓ(α, β; S=S, P=collect(1:nv(G)), t=t, k=k) - logdetΓ(α, β; S=S, P=setdiff(1:nv(G),L), t=t, k=k)));
return Ω;
end
function loss(L; getα=getα, getβ=getβ, logdet=false)
rL = getrL(L);
Ω = getΩ(getα(), getβ(), rL, L, logdet);
return Ω / length(L);
end
dat(x) = data.(data(x));
function call_back()
@printf("%6.3f, %6.3f, [%s], %6.3f\n",
accuracyFun(labels[L], dat(pred(L,V; G=G,labels=labels,predict=getRegression,α=((correlation == "homo") ? ones(length(A)) : getα()),β=getβ(),S=S))),
accuracyFun(labels[V], dat(pred(V,L; G=G,labels=labels,predict=getRegression,α=((correlation == "homo") ? ones(length(A)) : getα()),β=getβ(),S=S))),
array2str(getα()),
getβ());
end
mini_batch_size = Int(round(length(L) * 0.05));
mini_batches = [sample(L, mini_batch_size, replace=false) for _ in 1:num_steps];
train!(loss, getrL, getΩ,
(correlation == "learned") ? true : false,
θ,
params(ab),
mini_batches, L,
optθ, optφ;
cb=call_back, φ_start=0, φ_skip=φ_skip, cb_skip=100);
push!(rrt, accuracyFun(labels[U], dat(pred(U,L; G=G,labels=labels,predict=getRegression,α=((correlation == "homo") ? ones(length(A)) : getα()),β=getβ(),S=S))));
push!(αt, getα());
push!(βt, getβ());
@printf("\n pctg, rr, α, β\n");
@printf("%6.3f, %6.3f, [%s], %6.3f\n", ptr, rrt[end], array2str(getα()), getβ());
if inductive
if regressor == "gnn" && correlation == "zero"
@save "logs/gnn.bson" enc reg ab;
elseif regressor == "mlp" && correlation == "zero"
@save "logs/mlp.bson" mlp ab;
end
for i in 1:length(network_ind)
Random.seed!(seed_val);
G, A, labels, feats = read_network(network_ind[i]); n = nv(G);
d = sum(sum(A), dims=1)[:];
S = [spdiagm(0=>d.^-0.5)*A_*spdiagm(0=>d.^-0.5) for A_ in A];
@printf("\n pctg_inductive, rr, α, β\n");
for j in 1:length(ptr_inductive)
L, VU = rand_split(n, ptr_inductive[j]);
V, U = VU[1:div(length(VU),2)], VU[div(length(VU),2)+1:end];
if regressor in ["mlp", "gnn"] && correlation == "zero" && length(L) > 0
mini_batch_size = Int(round(length(L) * 0.05));
mini_batches = [sample(L, mini_batch_size, replace=false) for _ in 1:500];
if regressor == "mlp"
@load "logs/mlp.bson" mlp ab;
θ = params(mlp);
elseif regressor == "gnn"
@load "logs/gnn.bson" enc reg ab;
θ = params(enc, reg);
end
train!(loss, getrL, getΩ,
false,
θ,
params(ab),
mini_batches, L,
ADAM(0.0005), Descent(0.1);
cb=call_back, φ_start=0, φ_skip=φ_skip, cb_skip=100);
end
push!(rri[i][j], accuracyFun(labels[VU], dat(pred(VU,L; G=G,labels=labels,predict=getRegression,α=((correlation == "homo") ? ones(length(A)) : getα()),β=getβ(),S=S))));
@printf("%6.3f, %6.3f, [%s], %6.3f\n", ptr_inductive[j], rri[i][j][end], array2str(getα()), getβ());
end
end
end
end
for seed_val in 1:num_ave
run_once(seed_val);
end
@printf("\n trans: %s, %6.3f ± %6.3f\n", network_trans, mean(rrt), std(rrt));
@printf(" α, β: [%s] ± [%s], %6.3f ± %6.3f\n", array2str(mean(αt)), array2str(std(αt)), mean(βt), std(βt));
if inductive
for i in 1:length(network_ind)
@printf(" ind: %s, [%s] ± [%s]\n", network_ind[i], array2str(mean.(rri[i])), array2str(std.(rri[i])));
end
end
end