def applyRasterMap(data_nt, n_X=30):
import numpy as np
from rastermap import Rastermap
model = Rastermap(n_components=1, n_X=30, nPC=200, init='pca')
model = model.fit(data_nt)
isort_full = np.argsort(model.embedding[:, 0])
return isort_full
def ROI_selection(self, loaded=False):
self.colormat = np.zeros((0, 10, 3), dtype=np.int64)
lROI = len(self.Rselected)
if lROI > 0:
self.selected = np.array(
[item for sublist in self.Rselected for item in sublist])
self.colormat = np.concatenate(self.Rcolors, axis=0)
if not loaded:
print('yo')
if lROI > 4:
self.Ur = np.zeros((lROI, self.U.shape[1]),
dtype=np.float32)
ugood = np.zeros((lROI, )).astype(np.int32)
for r, rc in enumerate(self.Rselected):
if len(rc) > 0:
self.Ur[r, :] = self.U[rc, :].mean(axis=0)
ugood[r] = 1
ugood = ugood.astype(bool)
if ugood.sum() > 4:
model = Rastermap(n_components=1,
n_X=20,
init=np.arange(
0, ugood.sum()).astype(
np.int32)[:, np.newaxis])
y = model.fit_transform(self.Ur[ugood, :])
y = y.flatten()
y2 = np.zeros((lROI, ))
y2[(ugood).nonzero()[0]] = y
print(y2)
rsort = np.argsort(y2)
print(rsort)
roiorder = []
for r in self.ROIorder:
roiorder.append((rsort == r).nonzero()[0][0])
self.ROIorder = roiorder
self.ROIs = [self.ROIs[i] for i in rsort]
self.Rselected = [self.Rselected[i] for i in rsort]
self.Rcolors = [self.Rcolors[i] for i in rsort]
self.selected = np.array([
item for sublist in self.Rselected
for item in sublist
])
self.colormat = np.concatenate(self.Rcolors, axis=0)
else:
self.selected = np.argsort(self.embedding[:, 0])
self.colormat = 255 * np.ones(
(self.sp.shape[0], 10, 3), dtype=np.int32)
self.colormat_plot = self.colormat.copy()
self.plot_activity()
self.plot_colorbar()
self.win.show()
def getRastermapEmbeddings(PV, nX=50, nbin=50):
"""
Parameters
----------
PV : float32/float64
population vector: Time x Number of cells matrix
Returns
-------
isort : manifold embedding
Sfilt : sorted signals from all neurons.
"""
# raster map to find underlying embeddings
model = Rastermap(n_components=1, n_X=nX).fit(PV.T)
# the manifold embedding is in model.embedding
isort = np.argsort(model.embedding[:,0])
# sort by embedding and smooth over neurons
Sfilt = running_average(PV.T[isort, :], nbin)
Sfilt = zscore(Sfilt, axis=1)
return isort, Sfilt
iscell = iscell[:, 0].astype(np.bool)
else:
iscell = iscell.astype(np.bool)
if iscell.size == S.shape[0]:
S = S[iscell, :]
print('iscell found and used to select neurons')
print('size of rastermap matrix')
print(S.shape)
if len(S.shape) > 2:
S = S.mean(axis=-1)
S = zscore(S, axis=1)
ops['mode'] = 'basic'
model = Rastermap(n_components=ops['n_components'],
n_X=ops['n_X'],
nPC=ops['nPC'],
init=ops['init'],
alpha=ops['alpha'],
K=ops['K'],
constraints=ops['constraints'],
annealing=ops['annealing'])
if ops['end_time'] == -1:
ops['end_time'] = S.shape[1]
ops['start_time'] = 0
train_time = np.zeros((S.shape[1], )).astype(bool)
print(train_time.shape)
train_time[np.arange(ops['start_time'],
ops['end_time']).astype(int)] = 1
model.fit(S[:, train_time])
proc = {
'embedding': model.embedding,
'uv': [model.u, model.v],
'ops': ops,
def rastermap_comp(dff_path, rastermap_save_path):
dff = np.load(dff_path)
model = Rastermap(n_components=1, n_X=100).fit(dff)
np.save(rastermap_save_path, model.embedding[:, 0])
parser = argparse.ArgumentParser(description='spikes')
parser.add_argument('--S', default=[], type=str, help='spiking matrix')
parser.add_argument('--ops', default=[], type=str, help='options file')
parser.add_argument('--iscell', default=[], type=str, help='which cells to use')
args = parser.parse_args()
if len(args.S)>0:
S = np.load(args.S)
ops = np.load(args.ops)
ops = ops.item()
if len(args.iscell) > 0:
iscell = np.load(args.iscell)
if iscell.ndim > 1:
iscell = iscell[:, 0].astype(np.bool)
else:
iscell = iscell.astype(np.bool)
if iscell.size == S.shape[0]:
S = S[iscell, :]
print('iscell found and used to select neurons')
print(S.shape)
S = zscore(S,axis=1)
model = Rastermap(ops['n_components'], ops['n_X'], ops['n_Y'], ops['nPC'],
ops['sig_Y'], ops['init'], ops['alpha'], ops['K'])
model.fit(S)
proc = {'embedding': model.embedding, 'usv': [model.u, model.sv, model.v],
'cmap': model.cmap, 'A': model.A, 'ops': ops, 'filename': args.S}
basename, fname = os.path.split(args.S)
np.save(os.path.join(basename, 'embedding.npy'), proc)
else:
gui.run()