Skip to content

Commit

Permalink
Update Companion model with additional arguments, removing sentiment …
Browse files Browse the repository at this point in the history 
…loss
  • Loading branch information
@lthoang
lthoang committed Jul 31, 2024
1 parent 3ae5656 commit 6e304ce
Showing 1 changed file with 36 additions and 51 deletions.
87 changes: 36 additions & 51 deletions cornac/models/companion/recom_companion.pyx
View file
Original file line number Diff line number Diff line change
Expand Up @@ -88,6 +88,18 @@ class Companion(Recommender):
lambda_bpr: float, optional, default: 10.0
The regularization parameter for BPR.
lambda_p: float, optional, default: 10.0
The regularization parameter aspect ranking on item.
lambda_a: float, optional, default: 10.0
The regularization parameter for item ranking by aspect.
lambda_y: float, optional, default: 10.0
The regularization parameter for positive opinion ranking.
lambda_z: float, optional, default: 10.0
The regularization parameter for negative opinion ranking.
max_iter: int, optional, default: 200000
Maximum number of iterations for training.
Expand Down Expand Up @@ -137,8 +149,10 @@ class Companion(Recommender):
use_item_aspect_popularity=True,
enum_window=None,
lambda_reg=0.1,
lambda_s=0.1,
lambda_r=10,
lambda_p=10,
lambda_a=10,
lambda_y=10,
lambda_z=10,
lambda_bpr=10,
max_iter=200000,
lr=0.1,
Expand All @@ -164,8 +178,10 @@ class Companion(Recommender):
self.alpha = alpha
self.lambda_reg = lambda_reg
self.lambda_bpr = lambda_bpr
self.lambda_s = lambda_s
self.lambda_r = lambda_r
self.lambda_p = lambda_p
self.lambda_a = lambda_a
self.lambda_y = lambda_y
self.lambda_z = lambda_z
self.use_item_aspect_popularity = use_item_aspect_popularity
self.max_iter = max_iter
self.lr = lr
Expand Down Expand Up @@ -458,7 +474,7 @@ class Companion(Recommender):

with trange(self.max_iter, disable=not self.verbose) as progress:
for epoch in progress:
correct, skipped, sentiment_correct, sentiment_skipped, loss, bpr_loss, sentiment_loss = self._fit_mter(
correct, skipped, loss, bpr_loss = self._fit(
rng_pos, rng_neg,
rng_pos_uia, rng_neg_uia,
rng_pos_uiaop, rng_pos_uiaon,
Expand All @@ -483,8 +499,6 @@ class Companion(Recommender):
"bpr_loss": "%.2f" % (bpr_loss / self.n_bpr_samples),
"correct": "%.2f%%" % (100.0 * correct / (self.n_bpr_samples - skipped)),
"skipped": "%.2f%%" % (100.0 * skipped / self.n_bpr_samples),
"sentiment_loss": "%.2f" % (sentiment_loss / max(self.n_sentiment_samples, 1)),
"sentiment_correct": "%.2f%%" % (100.0 * sentiment_correct / max(self.n_sentiment_samples - sentiment_skipped, 1)),
})

if self.verbose:
Expand All @@ -495,7 +509,7 @@ class Companion(Recommender):
@cython.cdivision(True)
@cython.boundscheck(False)
@cython.wraparound(False)
def _fit_mter(
def _fit(
self,
RNGVector rng_pos,
RNGVector rng_neg,
Expand Down Expand Up @@ -547,8 +561,8 @@ class Companion(Recommender):
"""
cdef:
long s, i_index, j_index
long correct = 0, correct_uia = 0, correct_uia_i = 0, correct_uiaop = 0, correct_uiaon = 0, sentiment_correct = 0, aspect_correct = 0
long skipped = 0, skipped_uia = 0, skipped_uia_i = 0, skipped_uiaop = 0, skipped_uiaon = 0, sentiment_skipped = 0
long correct = 0, correct_uia = 0, correct_uia_i = 0, correct_uiaop = 0, correct_uiaon = 0, aspect_correct = 0
long skipped = 0, skipped_uia = 0, skipped_uia_i = 0, skipped_uiaop = 0, skipped_uiaon = 0
long n_users = self.train_set.num_users
long n_items = self.train_set.num_items
long n_aspects = self.train_set.sentiment.num_aspects
Expand All @@ -559,22 +573,22 @@ class Companion(Recommender):
long n_opinion_factors = self.n_opinion_factors
int num_samples = self.n_element_samples
int num_bpr_samples = self.n_bpr_samples
int num_sentiment_samples = self.n_sentiment_samples
int num_aspect_ranking_samples = self.n_aspect_ranking_samples
int num_opinion_ranking_samples = self.n_opinion_ranking_samples

integral _, i, j, k, idx, jdx, u_idx, i_idx, i_jdx, a_idx, a_jdx, o_idx, o_jdx, j_idx, thread_id
floating z, score, i_score, j_score, pred, temp, i_ij, uia_ij, uia_ij_i, uiaop_ij, uiaon_ij, a_ji
floating loss = 0., sentiment_loss = 0., bpr_loss = 0., bpr_loss_uia = 0., bpr_loss_uia_i = 0., bpr_loss_uiaop = 0., bpr_loss_uiaon = 0., aspect_bpr_loss = 0.
floating loss = 0., bpr_loss = 0., bpr_loss_uia = 0., bpr_loss_uia_i = 0., bpr_loss_uiaop = 0., bpr_loss_uiaon = 0., aspect_bpr_loss = 0.
floating del_sqerror, del_s_uia, del_bpr, del_bpr_uia, del_bpr_uia_i, del_bpr_uiaop, del_bpr_uiaon, del_aspect_bpr
floating eps = 1e-9

floating lr = self.lr
floating ld_reg = self.lambda_reg
floating ld_bpr = self.lambda_bpr
floating ld_s = self.lambda_s
floating ld_r = self.lambda_r

floating ld_p = self.lambda_p
floating ld_a = self.lambda_a
floating ld_y = self.lambda_y
floating ld_z = self.lambda_z
del_g1.fill(0)
del_g2.fill(0)
del_g3.fill(0)
Expand Down Expand Up @@ -610,7 +624,7 @@ class Companion(Recommender):
del_i[i_idx, j] += del_sqerror * G1[i, j, k] * U[u_idx, i] * A[a_idx, k]
del_a[a_idx, k] += del_sqerror * G1[i, j, k] * U[u_idx, i] * I[i_idx, j]

# get user aspect opinion element
# get user aspect positive opinion element
idx = rng_pos_uiaop.generate(thread_id)
u_idx = YP_uids[idx]
i_idx = YP_iids[idx]
Expand Down Expand Up @@ -641,7 +655,7 @@ class Companion(Recommender):
del_a[a_idx, j] += del_sqerror * G2[i, j, k] * I[i_idx, i] * O[o_idx, k]
del_o[o_idx, k] += del_sqerror * G2[i, j, k] * I[i_idx, i] * A[a_idx, j]

# get user aspect neg opinion element
# get user aspect negative opinion element
idx = rng_pos_uiaon.generate(thread_id)
u_idx = YN_uids[idx]
i_idx = YN_iids[idx]
Expand Down Expand Up @@ -673,35 +687,6 @@ class Companion(Recommender):
del_o[o_idx, k] += del_sqerror * G3[i, j, k] * I[i_idx, i] * A[a_idx, j]


for _ in prange(num_sentiment_samples, schedule='guided'):
idx = rng_pos_uia.generate(thread_id)
u_idx = X_uids[idx]
i_idx = X_iids[idx]
a_idx = X_aids[idx]
score = X[idx]
s = 1
if score < 3.0:
s = -1
elif score == 3.0:
sentiment_skipped += 1
continue

pred = get_score(G1, n_user_factors, n_item_factors, n_aspect_factors, U, I, A, u_idx, i_idx, a_idx)
pred = (pred - 3) * s
z = (1.0 / (1.0 + exp(pred)))
if z < .5:
sentiment_correct += 1
del_s_uia = ld_s * z * s
sentiment_loss += ld_s * log(1 / (1 + exp(-pred)))
for i in range(n_user_factors):
for j in range(n_item_factors):
for k in range(n_aspect_factors):
del_g1[i, j, k] -= del_s_uia * U[u_idx, i] * I[i_idx, j] * A[a_idx, k]
del_u[u_idx, i] -= del_s_uia * G1[i, j, k] * I[i_idx, j] * A[a_idx, k]
del_i[i_idx, j] -= del_s_uia * G1[i, j, k] * U[u_idx, i] * A[a_idx, k]
del_a[a_idx, k] -= del_s_uia * G1[i, j, k] * U[u_idx, i] * I[i_idx, j]


for _ in prange(num_bpr_samples, schedule='guided'):
idx = rng_pos.generate(thread_id)
u_idx = user_ids[idx]
Expand Down Expand Up @@ -765,7 +750,7 @@ class Companion(Recommender):
z = (1.0 / (1.0 + exp(pred)))
if z < .5:
correct_uia += 1
del_bpr_uia = ld_r * z * s
del_bpr_uia = ld_p * z * s

bpr_loss_uia += log(1 / (1 + exp(-pred)))
for k in range(n_aspect_factors):
Expand Down Expand Up @@ -802,7 +787,7 @@ class Companion(Recommender):
z = (1.0 / (1.0 + exp(pred)))
if z < .5:
correct_uia_i += 1
del_bpr_uia_i = ld_r * z * s
del_bpr_uia_i = ld_a * z * s

bpr_loss_uia_i += log(1 / (1 + exp(-pred)))
for j in range(n_item_factors):
Expand Down Expand Up @@ -841,7 +826,7 @@ class Companion(Recommender):
z = (1.0 / (1.0 + exp(pred)))
if z < .5:
correct_uiaop += 1
del_bpr_uiaop = ld_r * z * s
del_bpr_uiaop = ld_y * z * s

bpr_loss_uiaop += log(1 / (1 + exp(-pred)))
for k in range(n_opinion_factors):
Expand Down Expand Up @@ -886,7 +871,7 @@ class Companion(Recommender):
z = (1.0 / (1.0 + exp(pred)))
if z < .5:
correct_uiaon += 1
del_bpr_uiaon = ld_r * z * s
del_bpr_uiaon = ld_z * z * s

bpr_loss_uiaon += log(1 / (1 + exp(-pred)))
for k in range(n_opinion_factors):
Expand Down Expand Up @@ -982,7 +967,7 @@ class Companion(Recommender):
if O[i, j] < 0:
O[i, j] = 0

return correct, skipped, sentiment_correct, sentiment_skipped, loss, bpr_loss, sentiment_loss
return correct, skipped, loss, bpr_loss

def score(self, u_idx, i_idx=None):
"""Predict the scores/ratings of a user for an item.
Expand Down

0 comments on commit 6e304ce

Please sign in to comment.