-
Notifications
You must be signed in to change notification settings - Fork 3
/
Baseline_Models.m
283 lines (214 loc) · 6.83 KB
/
Baseline_Models.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
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
%% simple baseline predicition models
% scken, 2021
% Copyright (C) 2021 Chair of Automation Technology / TU Chemnitz
% parameter setup
dim = 576;
frac_scale = 6;
disp('----------------------------')
disp(['Baseline ' dataset])
disp(['Dim: ' num2str(dim)])
%% if dataset = full_crossval, create the training and test splits
if contains(dataset,'full_crossval')
%%% HDC encoding with kNN classifier
% load the data with the python script
ret = system(['python3 create_train_test_split_MATLAB.py --dataset=' dataset ' --preproc=1 --input_dim=' num2str(dim) ' --scale=' num2str(frac_scale)]);
if ret==0
load('temp_data.mat')
else
disp('Data could not converted')
return
end
delete('temp_data.mat')
for i=1:size(X_train,2)
%%%
% HDC with k-NN
% load data into item memory
VSA = vsa_env('vsa','FHRR','dim',dim);
VSA.add_vector('vec',X_train{i}','name',num2cell(num2str(Y_train{i})));
% find k nearest neigbors
tic
[~, l, s] = VSA.find_k_nearest(X_test{i}',3);
pred = [];
for c=1:size(l,2)
temp = str2num(cell2mat(l(:,c)));
pred(end+1) = mode(temp);
end
disp('Time for testing k-NN:')
toc
disp('Accuracy of HDC k-NN method: ')
f1 = getF1Score(Y_test{i},pred);
disp(f1)
end
%%%
% spectral features (FFT) with kNN
ret = system(['python3 create_train_test_split_MATLAB.py --dataset=' dataset ' --preproc=0 --input_dim=' num2str(dim) ' --scale=' num2str(frac_scale)]);
if ret==0
load('temp_data.mat')
else
disp('Data could not converted')
return
end
delete('temp_data.mat')
for i=1:size(X_train,2)
% fourier transformation
X_train{i} = abs(fft(X_train{i},size(X_train{i},2),2));
X_test{i} = abs(fft(X_test{i},size(X_test{i},2),2));
% concat input
X_train{i} = reshape(X_train{i},size(X_train{i},1),[]);
X_test{i} = reshape(X_test{i},size(X_test{i},1),[]);
Mdl = fitcknn(X_train{i},Y_train{i},'NumNeighbors',1,'Distance','Cityblock');
% testing
pred = predict(Mdl, X_test{i});
disp('Accuracy of Spectral Features kNN method: ')
f1 = getF1Score(Y_test{i},pred);
disp(f1)
end
end
%% HDC with SVM
ret = system(['python3 create_train_test_split_MATLAB.py --dataset=' dataset ' --preproc=1 --input_dim=' num2str(dim) ' --scale=' num2str(frac_scale)]);
if ret==0
load('temp_data.mat')
else
disp('Data could not converted')
return
end
delete('temp_data.mat')
tic
Mdl = fitcecoc(X_train,Y_train);
disp('Time for training HDC-SVM:')
toc
% testing
tic
pred = predict(Mdl, X_test);
disp('Time for testing HDC-SVM:')
toc
f1 = getF1Score(Y_test,pred);
disp('Accuracy of HDC SVM:')
disp(f1)
% add result to table
Result = table({'HDC-SVM'},f1,'VariableNames',{'Model','F1'});
%% HDC with k-NN
% load data into item memory
VSA = vsa_env('vsa','FHRR','dim',dim);
VSA.add_vector('vec',X_train','name',num2cell(num2str(Y_train)));
% find k nearest neigbors
tic
[~, l, s] = VSA.find_k_nearest(X_test',3);
pred = [];
for c=1:size(l,2)
temp = str2num(cell2mat(l(:,c)));
pred(end+1) = mode(temp);
end
disp('Time for testing k-NN:')
toc
disp('Accuracy of HDC k-NN method: ')
f1 = getF1Score(Y_test,pred);
disp(f1)
% add to table
% Result = table({'HDC-kNN'},acc,'VariableNames',{'Model','F1'});
Result.Model{end+1} = 'HDC-kNN';
Result.F1(end) = f1;
%% concat with SVM
ret = system(['python3 create_train_test_split_MATLAB.py --dataset=' dataset ' --preproc=0 --input_dim=' num2str(dim) ' --scale=' num2str(frac_scale)]);
if ret==0
load('temp_data.mat')
else
disp('Data could not converted')
return
end
delete('temp_data.mat')
% concat input
X_train = reshape(X_train,size(X_train,1),[]);
X_test = reshape(X_test,size(X_test,1),[]);
Mdl = fitcecoc(X_train,Y_train);
% testing
pred = predict(Mdl, X_test);
f1 = getF1Score(Y_test,pred);
disp('Accuracy of Concat SVM method: ')
disp(f1)
% add to table
Result.Model{end+1} = 'Concat-SVM';
Result.F1(end) = f1;
%% concat with kNN
% find optimal hyperparameter for concat model
% rng(0)
% Mdl_opt = fitcknn([X_train; X_test],[Y_train; Y_test],'OptimizeHyperparameters','auto',...
% 'HyperparameterOptimizationOptions',...
% struct('AcquisitionFunctionName','expected-improvement-plus'))
%
% Mdl = fitcknn(X_train,Y_train,'NumNeighbors',Mdl_opt.NumNeighbors,'Distance',Mdl_opt.Distance);
Mdl = fitcknn(X_train,Y_train,'NumNeighbors',3,'Distance','Cityblock');
% testing
pred = predict(Mdl, X_test);
f1 = getF1Score(Y_test,pred);
disp('Accuracy of Concat k-NN method: ')
disp(f1)
% add to table
Result.Model{end+1} = 'Concat-kNN';
Result.F1(end) = f1;
%% spectral features (FFT) with SVM
ret = system(['python3 create_train_test_split_MATLAB.py --dataset=' dataset ' --preproc=0 --input_dim=' num2str(dim) ' --scale=' num2str(frac_scale)]);
if ret==0
load('temp_data.mat')
else
disp('Data could not converted')
return
end
delete('temp_data.mat')
% fourier transformation
X_train = abs(fft(X_train,size(X_train,2),2));
X_test = abs(fft(X_test,size(X_test,2),2));
% concat input
X_train = reshape(X_train,size(X_train,1),[]);
X_test = reshape(X_test,size(X_test,1),[]);
tic
% Mdl = fitcecoc(X_train,Y_train,'Learners',svm_template);
Mdl = fitcecoc(X_train,Y_train);
disp('Time for training SVM-Stat:')
toc
% testing
tic
pred = predict(Mdl, X_test);
disp('Time for testing SVM-Stat:')
toc
disp('Accuracy of Spectral Features SVM method: ')
f1 = getF1Score(Y_test,pred);
disp(f1)
% add to table
Result.Model{end+1} = 'Spect-SVM';
Result.F1(end) = f1;
%%%
% spectral features (FFT) with kNN
ret = system(['python3 create_train_test_split_MATLAB.py --dataset=' dataset ' --preproc=0 --input_dim=' num2str(dim) ' --scale=' num2str(frac_scale)]);
if ret==0
load('temp_data.mat')
else
disp('Data could not converted')
return
end
delete('temp_data.mat')
% fourier transformation
X_train = abs(fft(X_train,size(X_train,2),2));
X_test = abs(fft(X_test,size(X_test,2),2));
% concat input
X_train = reshape(X_train,size(X_train,1),[]);
X_test = reshape(X_test,size(X_test,1),[]);
% find optimal hyperparameter for concat model
% rng(0)
% Mdl_opt = fitcknn([X_train; X_test],[Y_train; Y_test],'OptimizeHyperparameters','auto',...
% 'HyperparameterOptimizationOptions',...
% struct('AcquisitionFunctionName','expected-improvement-plus'))
%
% Mdl = fitcknn(X_train,Y_train,'NumNeighbors',Mdl_opt.NumNeighbors,'Distance',Mdl_opt.Distance);
Mdl = fitcknn(X_train,Y_train,'NumNeighbors',1,'Distance','Cityblock');
% testing
pred = predict(Mdl, X_test);
disp('Accuracy of Spectral Features kNN method: ')
f1 = getF1Score(Y_test,pred);
disp(f1)
% add to table
Result.Model{end+1} = 'Spect-kNN';
Result.F1(end) = f1;
%% print results
disp([dataset ' Dataset:'])
disp(Result)