def __init__(self, subjects=SUBJECTS, series=SERIES): training_ds = IOutils.data_streamer(patients_list=subjects, series_list=series) all_data = list(training_ds) X,Y = zip(*all_data) self.data = X[0] self.events = Y[0] self.mean = self.data.mean(axis=0) self.std = self.data.std(axis=0) self.normalize()
#IOutils.LABEL_NAMES has all the classes we wish to predict
#HI
# initialize logistic regressors
LRs = {}
LRsprocessed = {}
for label_name in IOutils.LABEL_NAMES:
# each label will have its own logistic regressor
LRs[label_name] = LogisticRegression()
LRsprocessed[label_name] = LogisticRegression()
print('Initialized logistic regressors')
# Load training data
# load 1 trial each from 3 patients
train_data = IOutils.data_streamer(mode='train', num_patients=1, num_series=8)
filters = ['alpha', 'beta']
# obtain a validation set
X_valid, Y_valid = train_data.next()
# X_valid = X_valid[]
selected_channels = range(X_valid.shape[1])
# selected_channels = [3,4]
X_valid = X_valid[:,selected_channels]
# print Y_valid
# X_valid = np.array(X_valid)
Y_valid = np.array(Y_valid)
]
# In[4]:
nn = NeuralNet(layers_list,
max_epochs=30,
update=nesterov_momentum,
update_learning_rate=0.02,
verbose=1000,
**LF.kwargs)
# In[5]:
training_ds = IOutils.data_streamer(patients_list=[2], series_list=range(1,7))
# nn = BasicCNN(input_shape=(None,42), output_num_units=12, max_epochs=50, hidden=[256, 120], add_drops=[1,1])
vt = IOutils.VectorTransformer()
# In[ ]:
n_repeat_sampling = 1
dataset_count = 0
for X,Y in training_ds:
X = X.astype(np.float)
X[np.isnan(X)] = 0
X = X/X.max()
wg = window_generator_ND(X, window_size=WINDOW_SIZE)
dataset_count += 1
# transform the Ys