def feat_rnn(c=0):
feats_eeg = scipy.stats.zscore(tools.feat_eeg(data[:,:,0]))
feats_emg = scipy.stats.zscore(tools.feat_emg(data[:,:,1]))
feats_eog = scipy.stats.zscore(tools.feat_eog(data[:,:,2]))
feats_all = np.hstack([feats_eeg, feats_eog, feats_emg])
feats_seq, targ_seq, groups_seq = tools.to_sequences(feats_all, target, groups=groups, seqlen=6, tolist=False)
r = cv(feats_seq, targ_seq, groups_seq, models.pure_rnn_do, name = 'feat-rnn-all', stop_after=15, counter=c, plot=plot)
with open('edfxresults_recurrent_feat.pkl', 'wb') as f: pickle.dump(r, f)
def feat_ann(c=0):
batch_size =700
feats_eeg = scipy.stats.zscore(tools.feat_eeg(data[:,:,0]))
feats_emg = scipy.stats.zscore(tools.feat_emg(data[:,:,1]))
feats_eog = scipy.stats.zscore(tools.feat_eog(data[:,:,2]))
feats_all = np.hstack([feats_eeg, feats_emg, feats_eog])
results = dict()
r = cv(feats_eeg, target, groups, models.ann, name = 'eeg', stop_after=15,batch_size=batch_size, counter=c, plot=plot)
results.update(r)
r = cv(np.hstack([feats_eeg,feats_eog]), target, groups, models.ann, name = 'eeg+eog',batch_size=batch_size, stop_after=15, counter=c, plot=plot)
results.update(r)
r = cv(np.hstack([feats_eeg,feats_emg]), target, groups, models.ann, name = 'eeg+emg',batch_size=batch_size, stop_after=15, counter=c, plot=plot)
results.update(r)
r = cv(feats_all, target, groups, models.ann, name = 'all',batch_size=batch_size, stop_after=15, counter=c, plot=plot)
results.update(r)
with open('results_electrodes_feat.pkl', 'wb') as f: pickle.dump(results, f)
def lstm(c=0):
batch_size = 256
name = 'CNN+LSTM'
rnn = {'model':models.pure_rnn_do, 'layers': ['fc1'], 'seqlen':6,
'epochs': 250, 'batch_size': 512, 'stop_after':15, 'balanced':False}
model = models.cnn3adam_filter_l2
r = cv (data, target, groups, model, rnn, name=name, batch_size=batch_size,
counter=counter, plot=plot, stop_after=15, balanced=False, cropsize=2800)
with open('results_recurrent_lstm.pkl', 'wb') as f: pickle.dump(r, f)
#%%
batch_size = 256
epochs = 256
###
rnn = {
'model': models.bi_lstm,
'layers': ['fc1'],
'seqlen': 6,
'epochs': 250,
'batch_size': 512,
'stop_after': 15,
'balanced': False
}
print(rnn)
model = models.cnn3adam_filter_morel2
results = keras_utils.cv(data,
target,
groups,
model,
rnn=rnn,
name='edfx-sample',
epochs=epochs,
folds=5,
batch_size=batch_size,
counter=0,
plot=True,
stop_after=15,
balanced=False,
cropsize=2800)
with open('results_dataset_edfx-sample.pkl', 'wb') as f:
pickle.dump(results, f)
# print(comment)
# plot = False
# ##%%
# epochs = 250
# batch_size = 512
#
results = dict()
#
# import stopping
# with open('results_electrodes.pkl', 'wb') as f:
# pickle.dump(results, f)
###%%
epochs = 250
batch_size = 256
#
cropsize = 2800
# r = cv(data[:,:,0:1], target, groups, models.cnn3adam_filter_morel2, epochs=epochs, name = 'eeg', stop_after=15, counter=counter,batch_size=batch_size, cropsize=cropsize)
# results.update(r)
## r = cv(data[:,:,[0,1]], target, groups, models.cnn3adam_filter_morel2, epochs=epochs, name = 'eeg+emg', stop_after=15, counter=counter,batch_size=batch_size, cropsize=cropsize)
# results.update(r)
# r = cv(data[:,:,[0,2]], target, groups, models.cnn3adam_filter_morel2, epochs=epochs, name = 'eeg+eog', stop_after=15, counter=counter,batch_size=batch_size, cropsize=cropsize)
# results.update(r)
r = cv(data[:,:,:], target, groups, models.cnn3dilated, epochs=epochs, name = 'all', stop_after=15, counter=counter,batch_size=batch_size, cropsize=cropsize, balanced=True)
# results.update(r)
# with open('results_electrodes_morel2.pkl', 'wb') as f:
# pickle.dump(results, f)
data, target, groups = load_data(dataset)
#%%
#s
batch_size = 256
epochs = 250
name = dataset
model = models.pure_rnn_do
data = tools.get_all_features(data)
feats = scipy.stats.zscore(data)
feats_seq, target_seq, groups_seq = tools.to_sequences(data,
target,
groups=groups,
seqlen=6)
results = keras_utils.cv(feats_seq,
target_seq,
groups_seq,
model,
name=name,
epochs=epochs,
folds=5,
batch_size=batch_size,
counter=counter,
plot=plot,
stop_after=15,
balanced=False)
with open('results_dataset_feat_{}'.format(dataset), 'wb') as f:
pickle.dump(results, f)
'balanced': False
}
print(rnn)
data = data[:12000]
target = target[:12000]
groups = groups[:12000]
# model = 'C:\\Users\\Simon\\dropbox\\Uni\\Masterthesis\\AutoSleepScorer\\weights\\balanced'
model = models.cnn3adam_filter_morel2
r = keras_utils.cv(data,
target,
groups,
model,
rnn=rnn,
trans_tuple=trans_tuple,
name=name,
epochs=epochs,
folds=5,
batch_size=batch_size,
counter=counter,
plot=True,
stop_after=15,
balanced=False,
cropsize=2800)
results.update(r)
with open('results_recurrejsuttestingnt_morel2.pkl', 'wb') as f:
pickle.dump(results, f)
telegram_send.send(parse_mode='Markdown',
messages=[
'DONE {} {}\n```\n{}\n```\n'.format(
os.path.basename(__file__), name,
tools.print_string(results))
groups = np.load('groups.npy')
feats_eeg = np.load('feats_eeg.npy')
feats_emg = np.load('feats_emg.npy')
feats_eog = np.load('feats_eog.npy')
feats_all = np.hstack([feats_eeg, feats_emg, feats_eog])
batch_size = 512
epochs = 250
balanced = True
results = {}
r = cv(feats_eeg,
target,
groups,
models.ann,
epochs=epochs,
name='eeg',
stop_after=15,
counter=counter,
batch_size=batch_size,
balanced=balanced)
results.update(r)
r = cv(np.hstack([feats_eeg, feats_emg]),
target,
groups,
models.ann,
epochs=epochs,
name='eeg+emg',
stop_after=15,
counter=counter,
batch_size=batch_size,
balanced=balanced)
def cnn_all(c=0):
r = cv(data[:,:,:], target, groups, models.cnn3adam_filter_l2, name = 'all', stop_after=15, counter=c, cropsize=2800, plot=plot)
with open('results_electrodes_cnn_all.pkl', 'wb') as f: pickle.dump(r, f)
target[target==5] = 4
target[target>5] = 0
target = keras.utils.to_categorical(target)
return data, target, groups
data,target,groups = load_data(dataset)
#%%
#s
# data = data[:15000]
# target = target[:15000]
# groups = groups[:15000]
batch_size = 256
epochs = 256
name = dataset
###
rnn = {'model':models.bi_lstm, 'layers': ['fc1'], 'seqlen':6,
'epochs': 250, 'batch_size': 512, 'stop_after':15, 'balanced':False}
print(rnn)
model = models.cnn3adam_filter_morel2
results = keras_utils.cv (data, target, groups, model, rnn=rnn, name=name,
epochs=epochs, folds=5, batch_size=batch_size, counter=counter,
plot=plot, stop_after=25, balanced=False, cropsize=2800)
with open('results_dataset_{}_zscore.pkl'.format(dataset), 'wb') as f:
pickle.dump(results, f)
# telegram_send.send(parse_mode='Markdown',messages=['DONE {} {} \n```\n{}\n```\n'.format(os.path.basename(__file__), dataset, tools.print_string(results))])
#
