Add RP3Beta

recbole/model/general_recommender/__init__.py

@@ -11,4 +11,5 @@
 from recbole.model.general_recommender.neumf import NeuMF
 from recbole.model.general_recommender.ngcf import NGCF
 from recbole.model.general_recommender.pop import Pop
+from recbole.model.general_recommender.rp3beta import RP3Beta
 from recbole.model.general_recommender.spectralcf import SpectralCF

recbole/model/general_recommender/rp3beta.py

@@ -0,0 +1,98 @@
+r"""
+RP3Beta
+################################################
+Reference:
+    Paudel, Bibek et al. Updatable, Accurate, Diverse, and Scalable Recommendations for Interactive Applications. https://doi.org/10.1145/2955101
+
+Reference code:
+    https://github.com/MaurizioFD/RecSys2019_DeepLearning_Evaluation/blob/master/GraphBased/RP3betaRecommender.py
+"""
+
+
+from recbole.utils.enum_type import ModelType
+import numpy as np
+import scipy.sparse as sp
+import torch
+
+from recbole.utils import InputType
+from recbole.model.abstract_recommender import GeneralRecommender
+
+
+def get_inv_degree_matrix(A):
+    # add epsilon to degree sums to suppress warning about division by zero when we later divide
+    degree_sums = A.sum(axis=1).getA1() + 1e-7
+
+    return sp.diags(1/degree_sums)
+
+
+# for reference, doing it in one computation
+# since the resultant matrix is dense, I'll refrain from doing this
+def calculate_rp3beta(B, beta):
+    user_degree_inv = get_inv_degree_matrix(B)
+    item_degree_inv = get_inv_degree_matrix(B.T)
+
+    # multiplication on left for row-wise scaling
+    user_transition = user_degree_inv @ B
+    item_transition = item_degree_inv @ B.T
+
+    P3 = user_transition @ item_transition @ user_transition
+
+    # multiplication on right for column-wise scaling (i.e., we're reweighting by inverse item popularity to a power)
+    RP3Beta = P3 @ item_degree_inv.power(beta)
+
+    return RP3Beta
+
+
+class RP3Beta(GeneralRecommender):
+    input_type = InputType.POINTWISE
+    type = ModelType.TRADITIONAL
+
+    def __init__(self, config, dataset):
+        super().__init__(config, dataset)
+
+        # need at least one param
+        self.dummy_param = torch.nn.Parameter(torch.zeros(1))
+
+        B = dataset.inter_matrix(
+            form='coo').astype(np.float32)
+
+        self.beta = config['beta']
+
+        user_degree_inv = get_inv_degree_matrix(B)
+        item_degree_inv = get_inv_degree_matrix(B.T)
+
+        self.user_transition = user_degree_inv @ B
+        self.item_transition = item_degree_inv @ B.T
+        self.item_degree_inv = item_degree_inv
+
+    def forward(self):
+        pass
+
+    def calculate_loss(self, interaction):
+        return torch.nn.Parameter(torch.zeros(1))
+
+    def predict(self, interaction):
+        user = interaction[self.USER_ID].cpu().numpy()
+        item = interaction[self.ITEM_ID].cpu().numpy()
+
+        specific_user_transitions = self.user_transition[user]
+
+        # make all item predictions for specified users
+        user_all_items = specific_user_transitions @ self.item_transition @ self.user_transition @ self.item_degree_inv.power(
+            self.beta)
+
+        # then narrow down to specific items
+        # without this copy(): "cannot set WRITEABLE flag..."
+        item_predictions = user_all_items[range(len(user)), item.copy()]
+
+        return torch.from_numpy(item_predictions.getA1())
+
+    def full_sort_predict(self, interaction):
+        user = interaction[self.USER_ID].cpu().numpy()
+
+        specific_user_transitions = self.user_transition[user]
+
+        item_predictions = specific_user_transitions @ self.item_transition @ self.user_transition @ self.item_degree_inv.power(
+            self.beta)
+
+        return torch.from_numpy(item_predictions.todense().getA1())

recbole/properties/model/RP3Beta.yaml

@@ -0,0 +1 @@
+beta: 0.55

run_test_example.py

@@ -134,6 +134,10 @@
         'model': 'LINE',
         'dataset': 'ml-100k',
     },
+    'Test RP3Beta': {
+        'model': 'RP3Beta',
+        'dataset': 'ml-100k',
+    },
 
     # Context-aware Recommendation
     'Test FM': {

tests/model/test_model_auto.py

@@ -116,6 +116,13 @@ def test_line(self):
         objective_function(config_dict=config_dict,
                            config_file_list=config_file_list, saved=False)
 
+    def test_rp3beta(self):
+        config_dict = {
+            'model': 'RP3Beta',
+        }
+        objective_function(config_dict=config_dict,
+                           config_file_list=config_file_list, saved=False)
+
 
 class TestContextRecommender(unittest.TestCase):
     # todo: more complex context information should be test, such as criteo dataset