def kmeansselect(self):
kmeans_mdl = Kmeans(self.data, num_bases=self._nsub)
kmeans_mdl.initialization()
kmeans_mdl.factorize()
# pick data samples closest to the centres
idx = dist.vq(kmeans_mdl.data, kmeans_mdl.W)
return idx
def kmeansselect(self):
kmeans_mdl = Kmeans(self.data, num_bases=self._nsub)
kmeans_mdl.initialization()
kmeans_mdl.factorize()
# pick data samples closest to the centres
idx = dist.vq(kmeans_mdl.data, kmeans_mdl.W)
return idx
def _map_w_to_data(self):
""" Return data points that are most similar to basis vectors W
"""
# assign W to the next best data sample
self._Wmapped_index = vq(self.data, self.W)
self.Wmapped = np.zeros(self.W.shape)
# do not directly assign, i.e. Wdist = self.data[:,sel]
# as self might be unsorted (in non ascending order)
# -> sorting sel would screw the matching to W if
# self.data is stored as a hdf5 table (see h5py)
for i, s in enumerate(self._Wmapped_index):
self.Wmapped[:,i] = self.data[:,s]
def _map_w_to_data(self):
""" Return data points that are most similar to basis vectors W
"""
# assign W to the next best data sample
self._Wmapped_index = vq(self.data, self.W)
self.Wmapped = np.zeros(self.W.shape)
# do not directly assign, i.e. Wdist = self.data[:,sel]
# as self might be unsorted (in non ascending order)
# -> sorting sel would screw the matching to W if
# self.data is stored as a hdf5 table (see h5py)
for i, s in enumerate(self._Wmapped_index):
self.Wmapped[:, i] = self.data[:, s]
def select_hull_points(data, n=3):
""" select data points for pairwise projections of the first n
dimensions """
# iterate over all projections and select data points
idx = np.array([])
# iterate over some pairwise combinations of dimensions
for i in combinations(range(n), 2):
# sample convex hull points in 2D projection
convex_hull_d = quickhull(data[i, :].T)
# get indices for convex hull data points
idx = np.append(idx, vq(data[i, :], convex_hull_d.T))
idx = np.unique(idx)
return np.int32(idx)
def select_hull_points(data, n=3):
""" select data points for pairwise projections of the first n
dimensions """
# iterate over all projections and select data points
idx = np.array([])
# iterate over some pairwise combinations of dimensions
for i in combinations(range(n), 2):
# sample convex hull points in 2D projection
convex_hull_d = quickhull(data[i, :].T)
# get indices for convex hull data points
idx = np.append(idx, vq(data[i, :], convex_hull_d.T))
idx = np.unique(idx)
return np.int32(idx)
def updateW(self): idx = self._subfunc() idx = np.sort(np.int32(idx)) mdl_small = self._mfmethod(self.data[:, idx], num_bases=self._num_bases, niter=self._niter, show_progress=self._show_progress, compW=True) # initialize W, H, and beta mdl_small.initialization() # determine W mdl_small.factorize() self.mdl = self._mfmethod(self.data[:, :], num_bases=self._num_bases , niter=self._niterH, show_progress=self._show_progress, compW=False) self.mdl.initialization() if self._mapW: # compute pairwise distances #distance = vq(self.data, self.W) _Wmapped_index = dist.vq(self.mdl.data, mdl_small.W) # do not directly assign, i.e. Wdist = self.data[:,sel] # as self might be unsorted (in non ascending order) # -> sorting sel would screw the matching to W if # self.data is stored as a hdf5 table (see h5py) for i,s in enumerate(_Wmapped_index): self.mdl.W[:,i] = self.mdl.data[:,s] else: self.mdl.W = np.copy(mdl_small.W)
def update_w(self):
idx = self._subfunc()
idx = np.sort(np.int32(idx))
mdl_small = self._mfmethod(self.data[:, idx],
num_bases=self._num_bases,
show_progress=self._show_progress,
compute_w=True)
# initialize W, H, and beta
mdl_small.initialization()
# determine W
mdl_small.factorize()
self.mdl = self._mfmethod(self.data[:, :],
num_bases=self._num_bases ,
show_progress=self._show_progress,
compute_w=False)
self.mdl.initialization()
if self._mapW:
# compute pairwise distances
#distance = vq(self.data, self.W)
_Wmapped_index = dist.vq(self.mdl.data, mdl_small.W)
# do not directly assign, i.e. Wdist = self.data[:,sel]
# as self might be unsorted (in non ascending order)
# -> sorting sel would screw the matching to W if
# self.data is stored as a hdf5 table (see h5py)
for i,s in enumerate(_Wmapped_index):
self.mdl.W[:,i] = self.mdl.data[:,s]
else:
self.mdl.W = np.copy(mdl_small.W)
def update_h(self):
# and assign samples to the best matching centers
self.assigned = dist.vq(self.W, self.data)
self.H = np.zeros(self.H.shape)
self.H[self.assigned, range(self._num_samples)] = 1.0
def update_h(self):
# and assign samples to the best matching centers
self.assigned = dist.vq(self.W, self.data)
self.H = np.zeros(self.H.shape)
self.H[self.assigned, range(self._num_samples)] = 1.0
