-
Notifications
You must be signed in to change notification settings - Fork 3
/
gpKernSetup.m
193 lines (134 loc) · 8.06 KB
/
gpKernSetup.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
function GPModel = gpKernSetup(GPModel, trainx, trainy)
%% decide what kernel to use
if ~isProperlySet(GPModel,'kern') || ~isProperlySet(GPModel.kern, 'type')
if isProperlySet(GPModel.Options, 'preComputedKern')
GPModel.kern.type = 'preComputedKern';
elseif isProperlySet(GPModel.Options,'kern')
GPModel.kern.type = GPModel.Options.kern;
elseif isProperlySet(GPModel.Options, 'hypIniMethod')
switch GPModel.Options.hypIniMethod
case 'PyHistMatchKernIni'
GPModel.kern.type = 'histIntKern';
case {'UnitIniScale', 'UnitScale', 'useUnitIniScale','DimWiseSTD'}
GPModel.kern.type = 'SQExpKern';
otherwise
error('unsupported kernel type');
end
else
GPModel.kern.type = 'SQExpKern';
warning('Unable to find clue about type of kernel to use, using %s', GPModel.kern.type);
end
end
%% initialize hyp
GPModel.Options.kern_name_str = GPModel.kern.type;
if ~isProperlySet(GPModel.kern, 'hyp')
switch GPModel.kern.type
case 'preComputedKern'
GPModel.Options.kern_name_str = [GPModel.Options.kern_name_str '-' GPModel.Options.kern];
if isProperlySet(GPModel.Options, 'hyp0')
GPModel.kern.hyp = GPModel.Options.hyp0;
else
if iscell(trainx)
std_trainy = std(trainy);
cc = arrayfun(@(jj) std_trainy/mean(diag(trainx{jj})), 1:length(trainx));
cc = reshape(cc,[],1);
else
cc = std(trainy)/mean(diag(trainx));
end
GPModel.kern.hyp = [mean(cc)/4; cc];
end
%GPModel.kern.K = trainx; %the precomputed K is passed in instead of trainx
GPModel.kern.extract_data_var = @(hyp) [];
GPModel.kern.extract_noise_var = @(hyp) hyp(1);
GPModel.kern.extract_kern_hyp = @(hyp) hyp(2:end);
GPModel.kern.pack_into_hyp = @(data_var, noise_var, kern_hyp) ...
[noise_var; reshape(kern_hyp,[],1)];
GPModel.kern.hyp_LB = GPModel.Options.pos_guard_bound * ones(size(GPModel.kern.hyp));
GPModel.kern.hyp_UB = inf(size(GPModel.kern.hyp));
case 'histIntKern'
if isProperlySet(GPModel.Options,'hyp0')
GPModel.kern.hyp = GPModel.Options.hyp0;
else
GPModel.kern.hyp = initializePyHistMatchHyp( [], ...
trainy,...
'PyHistMatchKernIni',...
GPModel.Options.dataSetName, ...
GPModel.Options.do_rand_corrupt_hyp_ini);
end
GPModel.kern.cellDataPrepFunc = @(data) histChannelMat2Cell(data, GPModel.Options.dataSetName);
Xcells = GPModel.kern.cellDataPrepFunc(trainx);
try
[~,HIQ] = histInterKern(@hist_isect_c, Xcells, Xcells,...
GPModel.kern.hyp(2:end));
GPModel.kern.KFunc_byInd = @(RI,CI,kern_hyp) ...
histInterKern(@hist_isect_c,...
[],...
[],...
kern_hyp,...
HIQ(RI,CI,:));
catch err
if ~strcmp(err.identifier, 'MATLAB:nomem')
rethrow(err);
end
end
GPModel.kern.KFunc = @(X0, X1, kern_hyp) histInterKern(@hist_isect_c, ...
X0, ...
X1, ...
kern_hyp);
GPModel.kern.diagKFunc = @(xx, nn, kern_hyp) sum(kern_hyp) * ones(nn,1);
GPModel.kern.dKFunc_dkernhyp = @(X0, X1, kern_hyp, dd, varargin) ...
histInterKern_dp(@hist_isect_c, ...
X0, ...
X1, ...
kern_hyp, ...
dd, ...
varargin{:});
GPModel.kern.extract_data_var = @(hyp) 1;
GPModel.kern.extract_noise_var = @(hyp)hyp(1);
GPModel.kern.extract_kern_hyp = @(hyp) hyp(2:end);
GPModel.kern.pack_into_hyp = @(data_var,noise_var,kern_hyp) ...
[noise_var; reshape(kern_hyp,[],1)];
GPModel.kern.hyp_LB = GPModel.Options.pos_guard_bound * ones(size(GPModel.kern.hyp));
GPModel.kern.hyp_UB = inf(size(GPModel.kern.hyp));
case 'SQExpKern'
if isProperlySet(GPModel.Options,'hyp0')
GPModel.kern.hyp = GPModel.Options.hyp0;
else
if isProperlySet(GPModel.Options, 'hypIniMethod')
GPModel.kern.hyp = initializeSQExpHyp(trainx,...
trainy, ...
GPModel.Options.hypIniMethod,...
GPModel.Options.do_rand_corrupt_hyp_ini);
else
GPModel.kern.hyp = initializeSQExpHyp(trainx, ...
trainy, ...
'',...
GPModel.Options.do_rand_corrupt_hyp_ini);
end
end
GPModel.kern.KFunc = @(X0, X1, kern_hyp) kern_sqexp(X0,...
X1,...
kern_hyp,...
[]);
GPModel.kern.diagKFunc = @(xx, nn, kern_hyp) ones(nn,1);
GPModel.kern.dKFunc_dkernhyp = @(X0, X1, kern_hyp, dd, varargin) ...
dkern_sqexp_dp(X0, ...
X1, ...
kern_hyp, ...
dd, ...
varargin{:});
GPModel.kern.extract_data_var = @(hyp) hyp(1);
GPModel.kern.extract_noise_var = @(hyp) hyp(2);
GPModel.kern.extract_kern_hyp = @(hyp) hyp(3:end);
GPModel.kern.pack_into_hyp = @(data_var, noise_var, kern_hyp) ...
[data_var; noise_var; reshape(kern_hyp,[],1)];
GPModel.kern.hyp_LB = GPModel.Options.pos_guard_bound * ones(size(GPModel.kern.hyp));
GPModel.kern.hyp_UB = inf(size(GPModel.kern.hyp));
otherwise
error('Unsupported kernel');
end
end
if ~isProperlySet(GPModel.Options, 'hyp0')
GPModel.Options.hyp0 = GPModel.kern.hyp;
end
end