| import tensorflow as tf |
|
|
|
|
| def bpr_loss(user_emb, pos_item_emb, neg_item_emb): |
| score = tf.reduce_sum(tf.multiply(user_emb, pos_item_emb), 1) - tf.reduce_sum(tf.multiply(user_emb, neg_item_emb), 1) |
| loss = -tf.reduce_sum(tf.log(tf.sigmoid(score) + 10e-8)) |
| return loss |
|
|
|
|
| def InfoNCE(view1, view2, temperature): |
| pos_score = tf.reduce_sum(tf.multiply(view1, view2), axis=1) |
| ttl_score = tf.matmul(view1, view2, transpose_a=False, transpose_b=True) |
| pos_score = tf.exp(pos_score / temperature) |
| ttl_score = tf.reduce_sum(tf.exp(ttl_score / temperature), axis=1) |
| cl_loss = -tf.reduce_sum(tf.log(pos_score / ttl_score)) |
| return cl_loss |
|
|
|
|
| |
| def ssm_loss(user_emb, pos_item_emb, neg_item_emb): |
| user_emb = tf.nn.l2_normalize(user_emb, 1) |
| pos_item_emb = tf.nn.l2_normalize(pos_item_emb, 1) |
| neg_item_emb = tf.nn.l2_normalize(neg_item_emb, 1) |
| pos_score = tf.reduce_sum(tf.multiply(user_emb, pos_item_emb), 1) |
| ttl_score = tf.matmul(user_emb, neg_item_emb, transpose_a=False, transpose_b=True) |
| ttl_score = tf.concat([tf.reshape(pos_score, (-1, 1)), ttl_score], axis=1) |
| pos_score = tf.exp(pos_score / 0.2) |
| ttl_score = tf.reduce_sum(tf.exp(ttl_score / 0.2), axis=1) |
| return -tf.reduce_mean(tf.log(pos_score / ttl_score)) |
|
|
