def _reconstruct_matrix(self, shifts=None, force=True):
if not self._matrix_reconstructed or force:
if shifts:
self._matrix_reconstructed = divisi2.reconstruct(self._U, self._S, self._V, shifts=shifts)
else:
self._matrix_reconstructed = divisi2.reconstruct(self._U, self._S, self._V)
return self._matrix_reconstructed
def _reconstruct_matrix(self, shifts=None, force=True):
if not self._matrix_reconstructed or force:
if shifts:
self._matrix_reconstructed = divisi2.reconstruct(self._U,
self._S,
self._V,
shifts=shifts)
else:
self._matrix_reconstructed = divisi2.reconstruct(
self._U, self._S, self._V)
return self._matrix_reconstructed
def test_cnet_blend():
from divisi2.blending import blend, blend_svd
matrix = divisi2.network.conceptnet_matrix('en')
isa = divisi2.network.filter_by_relation(matrix, 'IsA').squish().normalize_all()
atloc = divisi2.network.filter_by_relation(matrix, 'AtLocation').squish().normalize_all()
Uref, Sref, Vref = blend([isa, atloc]).svd(k=3)
U, S, V = blend_svd([isa, atloc], k=3)
rec_ref = divisi2.reconstruct(Uref, Sref, Vref)
rec_opt = divisi2.reconstruct(U, S, V)
# Check a random sampling of the items.
import random
for row in random.sample(rec_ref.row_labels, 50):
for col in random.sample(rec_ref.col_labels, 50):
assert np.allclose(rec_ref.entry_named(row, col),
rec_opt.entry_named(row, col))
def __init__(self, matrix, dimensionality):
concept_axes, axis_weights, feature_axes = matrix.svd(k=dimensionality)
self.predict = divisi2.reconstruct(concept_axes,
axis_weights,
feature_axes)
self.concept_sim = divisi2.reconstruct_activation(concept_axes,
axis_weights,
post_normalize=True)
self.feature_sim = divisi2.reconstruct_activation(feature_axes,
axis_weights,
post_normalize=True)
def learn_iter(mat):
print "Starting learning process..."
for i in range(NUM_LEARN_ITER):
user_mat, axis_weights, movie_mat = learn(mat)
# Reconstruct the learning matrix here.
for j in range(NUM_USERS):
for k in range(NUM_MOVIES):
mat[j, k] = divisi2.dot(user_mat[i,:], movie_mat[j,:])
print "Learning process complete."
start_time = time.time()
predictions = divisi2.reconstruct(user_mat, axis_weights, movie_mat)
print "Matrix reconstruction (elapsed time: %f s)." % (time.time() - start_time)
return predictions
def __init__(self, matrix_path=data_path+'feature_matrix_zh.smat'): # AnalogySpace A = divisi2.load(matrix_path) self.A = A.normalize_all() self.concept_axes, axis_weights, self.feature_axes = self.A.svd(k=100) self.sim = divisi2.reconstruct_similarity(\ self.concept_axes, axis_weights, post_normalize=False) self.predict = divisi2.reconstruct(\ self.concept_axes, axis_weights, self.feature_axes) # Fast spreading activation assoc = divisi2.load(data_path+'assoc_matrix_zh.smat') self.assoc = assoc.normalize_all() U, S, _ = self.assoc.svd(k=100) self.spread = divisi2.reconstruct_activation(U, S)