def test_norm(self, count_matrix_dataset, vectorizer):
"""Normalization test."""
_, x, y = count_matrix_dataset
v = vectorizer(norm="l2")
v.fit(x, y)
matrix = v.transform(x)
norm = sp_norm(matrix, axis=1)
np.testing.assert_allclose(np.mean(norm), 1.)
def norm(self):
if self.isnone:
return None
if self.mode == MODE_NP or (self.mode == MODE_SP
and self.type == 'vector'):
return np_norm(self.x)
if self.mode == MODE_TT:
return self.x.norm()
if self.mode == MODE_SP and self.type == 'matrix':
return sp_norm(self.x)
def compute_cosine_similarity(
sentence_vector: Union[np.ndarray, sp_sp.spmatrix],
sentence_vector2: Union[np.ndarray, sp_sp.spmatrix]) -> float:
"""
Params:
- sentence_vector: sentence vector
- sentence_vector2: sentence vector2
returns: similarity score of the two sentences
"""
score = sentence_vector @ (sentence_vector2.transpose())
if isinstance(sentence_vector, sp_sp.spmatrix) and \
isinstance(sentence_vector2, sp_sp.spmatrix):
score = score.todense()[0, 0]
nrm, nrm2 = sp_norm(sentence_vector), sp_norm(sentence_vector2)
score = score / (nrm * nrm2) if nrm * nrm2 > 0 else 0
else:
nrm, nrm2 = norm(sentence_vector), norm(sentence_vector2)
score = score / (nrm * nrm2) if nrm * nrm2 > 0 else 0
return score
def _get_error(self, rec, test_matrix, norm_ord=2):
total_error = 0
test_users = test_matrix.shape[0]
for i in range(test_users):
vec = test_matrix.getrow(i)
dropped_vec = self._drop_entries(vec, 0.25)
estimated_vec = rec.get_estimated_vector(dropped_vec)
total_error += sp_norm(vec - estimated_vec, ord=norm_ord,
axis=0)[0]
return total_error
from oppgave4c import oppgave4c
y4_c = oppgave4c()
E = 1.3 * 10**10
I = (0.3 * 0.03**3) / 12
f = -480 * 0.3 * 0.03 * 9.81
# Regner ut vektoren med lengde 10
y4_e = (f / (E * I)) * ones((10, 1))
forward_error = norm(y4_e - y4_c, ord=inf)
rel_forward_error = forward_error / norm(y4_e, ord=inf)
# Antar at den relative bakoverfeilen er 2^-52
rel_backwards_error = 2**(-52)
# Feilforstørringen er relativ foroverfeil delt på relativ bakoverfeil
error_magnification = rel_forward_error / rel_backwards_error
# Lager matrisen fra oppgave 2
A = lagA(10)
# Konverterer matrisen til en 'sparse matrix' for å inverse den
A = csc_matrix(A)
A_Invers = inv(A)
# Kondisjonstallet
cond_A = sp_norm(A, ord=inf) * sp_norm(A_Invers, ord=inf)
# Printer ut svarene
print("Foroverfeil: ", forward_error)
print("Relativ foroverfeil:", rel_forward_error)
print("Feilforstørring:", error_magnification)
print("Kondisjonstall for A:", cond_A)