implement bpr and bpr_max loss

models.py

@@ -41,7 +41,6 @@ class MetaLearner(nn.Module):
 
         return x.view(size[0], 1, 1, self.out_size)
 
-
 class EmbeddingLearner(nn.Module):
     def __init__(self):
         super(EmbeddingLearner, self).__init__()
@@ -52,21 +51,58 @@ class EmbeddingLearner(nn.Module):
         n_score = score[:, pos_num:]
         return p_score, n_score
 
-def bpr_loss(p_scores, n_values):
-    p1 = p_scores[:,0,None]
-    p2 = p_scores[:,1,None]
+def bpr_loss(p_scores, n_values,device):
+    ratio = int(n_values.shape[1] / p_scores.shape[1])
+    temp_pvalues = torch.tensor([]).cuda(device=device)
+    for i in range(p_scores.shape[1]):
+        temp_pvalues = torch.cat((temp_pvalues, p_scores[:, i, None].expand(-1, ratio)), dim=1)
+
+    d = torch.sub(temp_pvalues,n_values)
+    t = F.logsigmoid(d)
+    loss = -1 * (1.0/n_values.shape[1]) * t.sum(dim=1)
+    loss = loss.sum(dim=0)
+    return loss
+
+def bpr_max_loss(p_scores, n_values,device):
+    s = F.softmax(n_values,dim=1)
+    ratio = int(n_values.shape[1] / p_scores.shape[1])
+    temp_pvalues = torch.tensor([]).cuda(device=device)
+    for i in range(p_scores.shape[1]):
+        temp_pvalues = torch.cat((temp_pvalues,p_scores[:,i,None].expand(-1,ratio)),dim=1)
+
+    d = torch.sigmoid(torch.sub(temp_pvalues,n_values))
+    t = torch.mul(s,d)
+    loss = -1 * torch.log(t.sum(dim=1))
+    loss = loss.sum()
+    return loss
+
+def top_loss(p_scores, n_values):
+    p1 = p_scores[:, 0, None]
+    p2 = p_scores[:, 1, None]
 
     num_neg = n_values.shape[1]
-    half_index = int(num_neg/2)
+    half_index = int(num_neg / 2)
+
+    d1 = torch.sub(p1, n_values[:, 0:half_index])
+    d2 = torch.sub(p2, n_values[:, half_index:])
+    # print("d1 shape:",d1.shape)
 
-    d1 = torch.sub(p1,n_values[:,0:half_index])
-    d2 = torch.sub(p2,n_values[:,half_index:])
+    # print("add shape:",torch.cat((d1,d2),dim=1).shape)
+    t1 = torch.sigmoid(torch.cat((d1,d2),dim=1))
+    # print("t1 shape:",t1.shape)
+    t2 = torch.sigmoid(torch.pow(n_values,2))
+    # print("t2 shape:",t2.shape)
 
-    t = F.logsigmoid(torch.add(d1,d2))
+    t3 = torch.add(t1,t2)
+    # print("t3 shape:",t3.shape)
 
-    loss = (-1) * t.sum() / n_values.shape[1]
+    loss = t3.sum()
+    # print(loss.shape)
+    # loss /= (n_values.shape[1] * p_scores.shape[0])
+    loss /= n_values.shape[1]
     return loss
 
+
 class MetaTL(nn.Module):
     def __init__(self, itemnum, parameter):
         super(MetaTL, self).__init__()
@@ -81,7 +117,8 @@ class MetaTL(nn.Module):
                                                     num_hidden2=200, out_size=100, dropout_p=0)
 
         self.embedding_learner = EmbeddingLearner()
-        self.loss_func = nn.MarginRankingLoss(self.margin)
+        # self.loss_func = nn.MarginRankingLoss(self.margin)
+        self.loss_func = bpr_loss
 
         self.rel_q_sharing = dict()
 
@@ -118,9 +155,9 @@ class MetaTL(nn.Module):
 
             # sorted,indecies = torch.sort(n_score, descending=True,dim=1)
             # n_values = sorted[:,0:p_score.shape[1]]
-            n_values = n_score
-            loss = bpr_loss(p_score,n_values)
+
             # loss = self.loss_func(p_score, n_values, y)
+            loss = self.loss_func(p_score,n_score,device=self.device)
             loss.backward(retain_graph=True)
 
             grad_meta = rel.grad

trainer.py

@@ -28,7 +28,7 @@ class Trainer:
 
         self.optimizer = torch.optim.Adam(self.MetaTL.parameters(), self.learning_rate)
 
-            
+
     def rank_predict(self, data, x, ranks):
         # query_idx is the idx of positive score
         query_idx = x.shape[0] - 1
@@ -54,13 +54,13 @@ class Trainer:
             self.optimizer.zero_grad()
             p_score, n_score = self.MetaTL(task, iseval, curr_rel)
             y = torch.Tensor([1]).to(self.device)
-            loss = self.MetaTL.loss_func(p_score, n_score, y)
+            loss = self.MetaTL.loss_func(p_score, n_score,self.device)
             loss.backward()
             self.optimizer.step()
         elif curr_rel != '':
             p_score, n_score = self.MetaTL(task, iseval, curr_rel)
             y = torch.Tensor([1]).to(self.device)
-            loss = self.MetaTL.loss_func(p_score, n_score, y)
+            loss = self.MetaTL.loss_func(p_score, n_score,self.device)
         return loss, p_score, n_score
 
     def train(self):
@@ -74,12 +74,12 @@ class Trainer:
             # sample one batch from data_loader
             train_task, curr_rel = self.train_data_loader.next_batch()
             loss, _, _ = self.do_one_step(train_task, iseval=False, curr_rel=curr_rel)
-            # print the loss on specific epoch
-            # if e % self.print_epoch == 0:
-            #     loss_num = loss.item()
-            #     print("Epoch: {}\tLoss: {:.4f}".format(e, loss_num))
+
             # do evaluation on specific epoch
             if e % self.eval_epoch == 0 and e != 0:
+                loss_num = loss.item()
+                print("Epoch: {}\tLoss: {:.4f}".format(e, loss_num))
+
                 print('Epoch  {} Validating...'.format(e))
                 valid_data = self.eval(istest=False, epoch=e)
 
@@ -105,6 +105,7 @@ class Trainer:
 
         t = 0
         temp = dict()
+        total_loss = 0
         while True:
             # sample all the eval tasks
             eval_task, curr_rel = data_loader.next_one_on_eval()
@@ -113,7 +114,8 @@ class Trainer:
                 break
             t += 1
 
-            _, p_score, n_score = self.do_one_step(eval_task, iseval=True, curr_rel=curr_rel)
+            loss, p_score, n_score = self.do_one_step(eval_task, iseval=True, curr_rel=curr_rel)
+            total_loss += loss
 
             x = torch.cat([n_score, p_score], 1).squeeze()
 
@@ -132,12 +134,14 @@ class Trainer:
 
         
         if istest:
+            print("TEST: \t test_loss: ",total_loss.item())
             print("TEST: \tMRR: {:.3f}\tNDCG@10: {:.3f}\tNDCG@5: {:.3f}\tNDCG@1: {:.3f}\tHits@10: {:.3f}\tHits@5: {:.3f}\tHits@1: {:.3f}\r".format(
                     temp['MRR'], temp['NDCG@10'], temp['NDCG@5'], temp['NDCG@1'], temp['Hits@10'], temp['Hits@5'], temp['Hits@1']),"\n")
             with open('results.txt', 'a') as f:
                 f.writelines("TEST: \tMRR: {:.3f}\tNDCG@10: {:.3f}\tNDCG@5: {:.3f}\tNDCG@1: {:.3f}\tHits@10: {:.3f}\tHits@5: {:.3f}\tHits@1: {:.3f}\r\n\n".format(
                     temp['MRR'], temp['NDCG@10'], temp['NDCG@5'], temp['NDCG@1'], temp['Hits@10'], temp['Hits@5'], temp['Hits@1']))
         else:
+            print("VALID: \t validation_loss: ", total_loss.item())
             print("VALID: \tMRR: {:.3f}\tNDCG@10: {:.3f}\tNDCG@5: {:.3f}\tNDCG@1: {:.3f}\tHits@10: {:.3f}\tHits@5: {:.3f}\tHits@1: {:.3f}\r".format(
                     temp['MRR'], temp['NDCG@10'], temp['NDCG@5'], temp['NDCG@1'], temp['Hits@10'], temp['Hits@5'], temp['Hits@1']))
             with open("results.txt",'a') as f: