-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathGBN_Testing.m
executable file
·226 lines (213 loc) · 8.53 KB
/
GBN_Testing.m
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
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
function [Accuracy_all,ParaLocal] = GBN_Testing(X,ParaGlobal,Tcurrent,Para,c_jmean)
a0 = 0.01;
b0 = 0.01;
e0 = 1;
f0 = 1;
ac=1;
bc=1;
%%
[V,N] = size(X);
switch Para.DataType
case 'Positive'
[ii,jj,M]=find(X>eps); %Treat values smaller than eps as 0
iijj=find(X>eps);
Xmask = sparse(ii,jj,X(iijj),size(X,1),size(X,2));
case 'Binary'
[ii,jj,M] = find(X);
iijj=find(X);
case 'Count'
% [~,~,WS,DS,~,~]= PartitionX(X,101);
otherwise
error('Wrong Input Datatype!');
end
%% Initial ParaGlobal
if strcmp(Para.DataType, 'Positive')
c = ParaGlobal.c;
if length(c)>1
c = mean(c)*ones(1,N);
end
end
Phi = ParaGlobal.Phi;
r_k = ParaGlobal.r_k;
% gamma0 = ParaGlobal.gamma0;
% c0 = ParaGlobal.c0;
%% Initial ParaLocal
c_j = cell(Tcurrent+1,1);
for t=1:Tcurrent+1
c_j{t}=ones(1,N)*c_jmean(t);
end
p_j = Calculate_pj(c_j,Tcurrent);
Theta = cell(Tcurrent,1);
for t=Tcurrent:-1:1
if t==Tcurrent
shape = r_k*ones(1,N);
else
if size(Phi{t+1},2)~=size(Theta{t+1},1)
save('GBNTesting_wrong.mat','Phi','Theta');
end
shape = Phi{t+1}*Theta{t+1};
end
Theta{t} = max( bsxfun(@times, randg(shape) , 1./c_j{t+1}) , 1e-2 );
end
%% Initialization
%ThetaAver = cell(Tcurrent,1);
ThetaFreqAver = cell(Tcurrent,1);
c_jAver = cell(Tcurrent+1,1);
p_jAver = cell(Tcurrent+1,1);
for t = 1:(Tcurrent+1)
if t <= Tcurrent
ThetaAver{t}=0;
ThetaFreqAver{t}=0;
end
c_jAver{t}=0;
p_jAver{t}=0;
end
Accuracy_all.Accuracy_Theta = [];
Xt_to_t1=cell(Tcurrent,1);
Accuracy_all.LogLikelihood =0;
%Accuracy_all.LogLikelihood_CSL=0;
%Accuracy_all.LogLikelihood_Harmonic=0;
Xmask=sparse(X);
%% Sample ParaLocal
for iter = 1 : (Para.TestBurnin + Para.TestCollection)
tic
%%==================================== Upward Pass ===================================
for t = 1:Tcurrent
if t == 1 %&& Tcurrent==1
switch Para.DataType
case 'Positive'
if length(c)==1
if Para.ParallelProcessing
Rate = mtimes_par(Phi{1},Theta{1}, Para.NumBlockParallel);
else
Rate = Phi{1}*Theta{1};
end
Rate = 2*sqrt(c*X(iijj).*Rate(iijj));
if Para.ParallelProcessing
M = Truncated_bessel_rnd_par( Rate, Para.NumBlockParallel);
else
M = Truncated_bessel_rnd( Rate );
end
else
if Para.ParallelProcessing
Rate = mtimes_par(Phi{1},Theta{1}, Para.NumBlockParallel);
else
Rate = Phi{1}*Theta{1};
end
Rate = 2*sqrt(c(jj)'.*X(iijj).*Rate(iijj));
if Para.ParallelProcessing
M = Truncated_bessel_rnd_par( Rate, Para.NumBlockParallel);
else
M = Truncated_bessel_rnd( Rate );
end
c = randg(full(sparse(1,jj,M,1,N))+ac) ./ (bc+sum(X,1));
end
Xt = sparse(ii,jj,M,V,N);
case 'Binary'
if Para.ParallelProcessing
Rate = Mult_Sparse_par(Xmask,Phi{1},Theta{1}, Para.NumBlockParallel);
else
Rate = Mult_Sparse(Xmask,Phi{1},Theta{1});
end
if Para.ParallelProcessing
M = Truncated_Poisson_rnd_par(full(Rate(iijj)), Para.NumBlockParallel);
else
M = truncated_Poisson_rnd(full(Rate(iijj)));
end
Xt = sparse(ii,jj,M,V,N);
case 'Count'
Xt = sparse(X);
end
if Para.ParallelProcessing
Xt_to_t1{t} = Multrnd_Matrix_mex_fast_v1_par(Xt,Phi{t},Theta{t}, Para.NumBlockParallel);
else
Xt_to_t1{t} = Multrnd_Matrix_mex_fast_v1(Xt,Phi{t},Theta{t});
end
else
if Para.ParallelProcessing
Xt_to_t1{t} = CRT_Multrnd_Matrix_par(sparse(Xt_to_t1{t-1}),Phi{t},Theta{t}, Para.NumBlockParallel);
else
Xt_to_t1{t} = CRT_Multrnd_Matrix(sparse(Xt_to_t1{t-1}),Phi{t},Theta{t});
end
end
end
%%==================================== Downward Pass ===================================
%%==================== Sample Theta ========================
if iter>10
if Tcurrent > 1
p_j{2} = betarnd( sum(Xt_to_t1{1},1)+a0 , sum(Theta{2},1)+b0 );
else
p_j{2} = betarnd( sum(Xt_to_t1{1},1)+a0 , sum(r_k,1)+b0 );
end
p_j{2} = min( max(p_j{2},realmin) , 1-realmin);
c_j{2} = (1-p_j{2})./p_j{2};
for t = 3:(Tcurrent+1)
if t == Tcurrent+1
c_j{t} = randg(sum(r_k)*ones(1,N)+e0) ./ (sum(Theta{t-1},1)+f0);
else
c_j{t} = randg(sum(Theta{t},1)+e0) ./ (sum(Theta{t-1},1)+f0);
end
end
p_j_temp = Calculate_pj(c_j,Tcurrent);
p_j(3:end)=p_j_temp(3:end);
end
for t = Tcurrent:-1:1
if t == Tcurrent
shape = r_k;
else
shape = Phi{t+1}*Theta{t+1};
end
if Para.ParallelProcessing
temp = randg_par(bsxfun(@plus,shape,Xt_to_t1{t}), Para.NumBlockParallel);
Theta{t} = bsxfun(@rdivide, temp , c_j{t+1}-log(max(1-p_j{t},realmin)) );
else
Theta{t} = bsxfun(@times, randg(bsxfun(@plus,shape,Xt_to_t1{t})), 1 ./ (c_j{t+1}-log(max(1-p_j{t},realmin))) );
end
if nnz(isnan(Theta{t}))
warning('Theta Nan');
Theta{t}(isnan(Theta{t}))=0;
end
end
Timetmp = toc;
if mod(iter,10)==0
fprintf('Testing Layer: %d, iter: %d, TimePerIter: %d seconds. \n',Tcurrent,iter,Timetmp);
end
%%==================================== Average ===================================
if iter > Para.TestBurnin %&& (mod(iter-Para.TestBurnin,TestSampleSpace)==0)
for t = 1:(Tcurrent+1)
if t <= Tcurrent
%ThetaAver{t} = ThetaAver{t} + Theta{t} / Para.TestCollection;
ThetaFreqAver{t} = ThetaFreqAver{t} + bsxfun(@rdivide,Theta{t},max(sum(Theta{t},1),realmin)) / Para.TestCollection;
end
c_jAver{t} = c_jAver{t} + c_j{t} / Para.TestCollection;
p_jAver{t} = p_jAver{t} + p_j{t} / Para.TestCollection;
end
% if strcmp(DataType, 'Binary')
% Rate1=Phi{1}*Theta{1}(:,prepar.teindx);
% PBinary = 1-exp(-Rate1);
% tmp = X(:,prepar.teindx) .* log(max(realmin, PBinary) ) + (1-X(:,prepar.teindx)).*(-Rate1);
% LogLikelihood = sum(tmp,1);
% Accuracy_all.LogLikelihood_CSL = Accuracy_all.LogLikelihood_CSL + exp(LogLikelihood)/TestCollection;
% Accuracy_all.LogLikelihood_Harmonic = Accuracy_all.LogLikelihood_Harmonic + exp(-LogLikelihood)/TestCollection;
% end
end
end %%======================= One Testing iteration End ===========================
%%============== Calculate Accuracy ========================
if ~Para.ParallelProcessing
%[AccuracyTmp] = TestingInside(ThetaFreqAver{1},Para);
Accuracy_all = TestingInside(ThetaFreqAver{1},Para);
else
%[AccuracyTmp] = TestingInside(ThetaFreqAver{1},Para,true,3);
Accuracy_all = TestingInside(ThetaFreqAver{1},Para,true,3);
end
%Accuracy_all.Accuracy_ThetaFreqAver = AccuracyTmp.Accuracy_ThetaFreq_accuracy(1);
%Accuracy_all.Accuracy_ThetaFreqAver_AccAver = AccuracyTmp.Accuracy_ThetaFreq_AccAver;
% if strcmp(DataType, 'Binary')
% Accuracy_all.LogLikelihood_CSL = mean(log(max(Accuracy_all.LogLikelihood_CSL,realmin)));
% Accuracy_all.LogLikelihood_Harmonic = mean(log(1./Accuracy_all.LogLikelihood_Harmonic));
% end
%%====== Save Local Parameters =============
%ParaLocal.ThetaAver = ThetaAver;
%ParaLocal.ThetaFreqAver = ThetaFreqAver;
ParaLocal.c_jAver = c_jAver;
ParaLocal.p_jAver = p_jAver;