def get_connections(site1, site2):
''' return mask of rows which are fragments common to both sites '''
return (~np.isnan(site1)) & (~np.isnan(site2))
def get_best_f1_ind(precision, recall):
fscore = (2 * precision * recall) / (precision + recall)
return np.argmax(fscore)
if __name__ == '__main__':
# load preprocessed data
fragments, qualities, variant_labels = load_preprocessed(
'data/preprocessed/chr20_1-1M.npz'
)
matrix_sparsity_info(fragments, print_info=True)
# get undirected edge weights
min_connections = 2
edges = np.ma.masked_all((fragments.shape[1], fragments.shape[1]))
# to catch places where phi undefined and therefore should be no edge
with warnings.catch_warnings():
warnings.simplefilter("error")
for i, j in tqdm(zip(*np.tril_indices(fragments.shape[1], k=-1)),
desc='edges',
total=((fragments.shape[1] ** 2) / 2 - fragments.shape[1])):
if thresh_ind < len(thresholds):
thresh = -thresholds[thresh_ind]
else:
thresh = 0
preds = scores > thresh
print(preds.sum())
return confusion_matrix(variant_labels, preds,
normalize=normalize), None
if __name__ == '__main__':
# load preprocessed data
_, _, variant_labels = load_preprocessed(
'data/preprocessed/chr20_1-1M.npz')
save_dir = 'data/results'
# original
res_saves = [
('phi_correlation_1M.npy', 'phi correlation'),
('likelihood_w1_1M.npy', 'likelihood w=1'),
('likelihood_w2_1M.npy', 'likelihood w=2'),
('likelihood_w3_1M.npy', 'likelihood w=3'),
('likelihood_w4_1M.npy', 'likelihood w=4'),
('likelihood_w5_1M.npy', 'likelihood w=5'),
]
results = [(np.load(os.path.join(save_dir, s[0])), s[1])
for s in res_saves]
parser.add_argument('-n',
'--n_splits',
type=int,
action='store',
default=10)
return parser
if __name__ == "__main__":
scaler = None
parser = gen_arg_parser()
args = parser.parse_args()
merged_data = load_preprocessed(args.data_dir_root,
simple=args.merge_nrem,
merge_keys=['stages', 'pows'])
if args.scaler:
scaler = StandardScaler()
scaler.fit(merged_data['pows'])
if args.mode == 'merged':
data = merged_data
else:
# Implement this
pass
if args.merge_nrem == True:
sleep_stages = data['stages_simple']
else:
sleep_stages = data['stages'][:, 2].astype(np.int8)
import os
import numpy as np
import scipy.io as sio
from sklearn.preprocessing import StandardScaler
from nn_models import basic_rnn
from utils import rolling_window, load_preprocessed
# Load preprocessed files
all_data = load_preprocessed()
data = all_data[0]
def timestep_slice_data(data, slice_size=10, rescale=True):
# Load inputs and outputs
labels = data['stages'][:, 2]
pows = data['pows']
if rescale:
scaler = StandardScaler()
scaler.fit(pows)
pows = scaler.transform(pows)
pows = pows.swapaxes(0, 1)
# timeslice labels [ N,slice_size ]
seq_labels = rolling_window(labels, slice_size)
# timeslicing pows is awkward...
seq_pows = rolling_window(pows, slice_size)
seq_pows = seq_pows.swapaxes(0, 1)
seq_pows = seq_pows.swapaxes(1, 2)