def _compute_similarity_matrix(self, knn):
for e in range(self.epochs):
self.iteration()
s_tmp = []
for i in tqdm(range(self.n_items)):
mat = self.sm[i, :]
s_tmp.append(utils.knn(mat, knn))
s = sparse.vstack(s_tmp, format='csr')
s.setdiag(0)
self.sm = s
def _compute_similarity_matrix(self, icm, knn):
s_tmp = []
n_items = icm.shape[0]
m = icm.tocsr()
m_t = m.T.tocsr()
for i in tqdm(range(n_items)):
mat = m[i, :].dot(m_t)
s_tmp.append(utils.knn(mat, knn))
s = sparse.vstack(s_tmp, format='csr')
s.setdiag(0)
return s
def _compute_similarity_matrix(self, knn):
s_tmp = []
user_factors, item_factors = self.ALS()
n_items = item_factors.shape[0]
item_factors = sparse.csr_matrix(item_factors)
item_factors_T = item_factors.T
for i in tqdm(range(n_items)):
mat = item_factors[i, :].dot(item_factors_T)
s_tmp.append(utils.knn(mat, knn))
s = sparse.vstack(s_tmp, format='csr')
s.setdiag(0)
return s
def _compute_similarity_matrix(self, icm, knn):
s_tmp = []
ucm = self.urm.T
n_items = icm.shape[0]
m1 = icm.tocsr()
m1_t = m1.T.tocsr()
m2 = ucm.tocsr()
m2_t = m2.T.tocsr()
for i in tqdm(range(n_items)):
cfb_i = m1[i, :].dot(m1_t)
cf_i = m2[i, :].dot(m2_t)
mat = self.cbf_weight*cfb_i + self.cf_weight*cf_i
s_tmp.append(utils.knn(mat, knn))
s = sparse.vstack(s_tmp, format='csr')
s.setdiag(0)
return s
