add comirec model

add comirec model

.github/ISSUE_TEMPLATE/bug_report.md

@@ -20,7 +20,7 @@ Steps to reproduce the behavior:
 **Operating environment(运行环境):**
  - python version [e.g. 3.6, 3.7, 3.8]
  - tensorflow version [e.g. 1.9.0, 1.14.0, 2.5.0]
- - deepmatch version [e.g. 0.3.0,]
+ - deepmatch version [e.g. 0.3.1,]
 
 **Additional context**
 Add any other context about the problem here.

.github/ISSUE_TEMPLATE/question.md

@@ -17,4 +17,4 @@ Add any other context about the problem here.
 **Operating environment(运行环境):**
  - python version [e.g. 3.6, 3.7, 3.8]
  - tensorflow version [e.g. 1.9.0, 1.14.0, 2.5.0]
- - deepmatch version [e.g. 0.3.0,]
+ - deepmatch version [e.g. 0.3.1,]

.github/workflows/ci.yml

@@ -17,7 +17,7 @@ jobs:
     timeout-minutes: 120
     strategy:
       matrix:
-        python-version: [3.6,3.7,3.8]
+        python-version: [3.6,3.7,3.8,3.9,3.10.7]
         tf-version: [1.9.0,1.14.0,2.5.0]
 
         exclude:
@@ -57,12 +57,28 @@ jobs:
             tf-version: 2.8.0
           - python-version: 3.6
             tf-version: 2.9.0
+          - python-version: 3.6
+            tf-version: 2.10.0
           - python-version: 3.9
             tf-version: 1.4.0
+          - python-version: 3.9
+            tf-version: 1.9.0
           - python-version: 3.9
             tf-version: 1.15.0
           - python-version: 3.9
-            tf-version: 2.2.0
+            tf-version: 1.14.0
+          - python-version: 3.10.7
+            tf-version: 1.4.0
+          - python-version: 3.10.7
+            tf-version: 1.9.0
+          - python-version: 3.10.7
+            tf-version: 1.15.0
+          - python-version: 3.10.7
+            tf-version: 1.14.0
+          - python-version: 3.10.7
+            tf-version: 2.5.0
+          - python-version: 3.10.7
+            tf-version: 2.6.0
 
     steps:
 
@@ -75,6 +91,7 @@ jobs:
 
     - name: Install dependencies
       run: |
+        sudo apt update && sudo apt install -y pkg-config libhdf5-dev
         pip3 install -q tensorflow==${{ matrix.tf-version }}
         pip install -q protobuf==3.19.0
         pip install -q requests

.gitignore

@@ -1,3 +1,4 @@
+.idea
 # Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]

README.md

@@ -32,6 +32,7 @@ Let's [**Get Started!**](https://deepmatch.readthedocs.io/en/latest/Quick-Start.
 | NCF  | [WWW 2017][Neural Collaborative Filtering](https://arxiv.org/abs/1708.05031)       |
 | SDM  | [CIKM 2019][SDM: Sequential Deep Matching Model for Online Large-scale Recommender System](https://arxiv.org/abs/1909.00385)  |
 | MIND | [CIKM 2019][Multi-interest network with dynamic routing for recommendation at Tmall](https://arxiv.org/pdf/1904.08030)  |
+| COMIREC | [KDD 2020][Controllable Multi-Interest Framework for Recommendation](https://arxiv.org/pdf/2005.09347.pdf)  |
 
 ## Contributors([welcome to join us!](./CONTRIBUTING.md))
 
@@ -60,6 +61,11 @@ Let's [**Get Started!**](https://deepmatch.readthedocs.io/en/latest/Quick-Start.
          <a href="https://github.com/LeoCai">LeoCai</a>
          <p>  ByteDance   </p>​
       </td>
+      <td>
+        ​ <a href="https://github.com/liyuan97"><img width="70" height="70" src="https://github.com/liyuan97.png?s=40" alt="pic"></a><br>
+        ​ <a href="https://github.com/liyuan97">Li Yuan</a>
+        <p> Tencent    </p>​
+      </td>
       <td>
         ​ <a href="https://github.com/yangjieyu"><img width="70" height="70" src="https://github.com/yangjieyu.png?s=40" alt="pic"></a><br>
         ​ <a href="https://github.com/yangjieyu">Yang Jieyu</a>

deepmatch/__init__.py

@@ -1,4 +1,4 @@
 from .utils import check_version
 
-__version__ = '0.3.0'
+__version__ = '0.3.1'
 check_version(__version__)

deepmatch/models/__init__.py

@@ -4,3 +4,4 @@
 from .ncf import NCF
 from .mind import MIND
 from .sdm import SDM
+from .comirec import ComiRec

deepmatch/models/comirec.py

@@ -0,0 +1,186 @@
+"""
+Author:
+    Li Yuan, lysysu@qq.com
+
+Reference:
+Yukuo Cen, Jianwei Zhang, Xu Zou, et al. Controllable Multi-Interest Framework for Recommendation//Accepted to KDD 2020
+"""
+
+import tensorflow as tf
+from deepctr.feature_column import SparseFeat, VarLenSparseFeat, DenseFeat, \
+    embedding_lookup, varlen_embedding_lookup, get_varlen_pooling_list, get_dense_input, build_input_features
+from deepctr.layers import DNN, PositionEncoding
+from deepctr.layers.utils import NoMask, combined_dnn_input, add_func
+from tensorflow.python.keras.layers import Concatenate, Lambda
+from tensorflow.python.keras.models import Model
+
+from ..inputs import create_embedding_matrix
+from ..layers.core import CapsuleLayer, PoolingLayer, MaskUserEmbedding, LabelAwareAttention, SampledSoftmaxLayer, \
+    EmbeddingIndex
+from ..layers.interaction import SoftmaxWeightedSum
+from ..utils import get_item_embedding
+
+
+def tile_user_otherfeat(user_other_feature, k_max):
+    return tf.tile(tf.expand_dims(user_other_feature, -2), [1, k_max, 1])
+
+
+def tile_user_his_mask(hist_len, seq_max_len, k_max):
+    return tf.tile(tf.sequence_mask(hist_len, seq_max_len), [1, k_max, 1])
+
+
+def softmax_Weighted_Sum(input):
+    history_emb_add_pos, mask, attn = input[0], input[1], input[2]
+    attn = tf.transpose(attn, [0, 2, 1])
+    pad = tf.ones_like(mask, dtype=tf.float32) * (-2 ** 32 + 1)
+    attn = tf.where(mask, attn, pad)  # [batch_size, seq_len, num_interests]
+    attn = tf.nn.softmax(attn)  # [batch_size, seq_len, num_interests]
+    high_capsule = tf.matmul(attn, history_emb_add_pos)
+    return high_capsule
+
+
+def ComiRec(user_feature_columns, item_feature_columns, k_max=2, p=100, interest_extractor='sa',
+            add_pos=True,
+            user_dnn_hidden_units=(64, 32), dnn_activation='relu', dnn_use_bn=False, l2_reg_dnn=0,
+            l2_reg_embedding=1e-6,
+            dnn_dropout=0, output_activation='linear', sampler_config=None, seed=1024):
+    """Instantiates the ComiRec Model architecture.
+
+    :param user_feature_columns: An iterable containing user's features used by  the model.
+    :param item_feature_columns: An iterable containing item's features used by  the model.
+    :param k_max: int, the max size of user interest embedding
+    :param p: float,the parameter for adjusting the attention distribution in LabelAwareAttention.
+    :param interest_extractor: string, type of a multi-interest extraction module, 'sa' means self-attentive and 'dr' means dynamic routing
+    :param add_pos: bool. Whether use positional encoding layer
+    :param user_dnn_hidden_units: list,list of positive integer or empty list, the layer number and units in each layer of user tower
+    :param dnn_activation: Activation function to use in deep net
+    :param dnn_use_bn: bool. Whether use BatchNormalization before activation or not in deep net
+    :param l2_reg_dnn:  L2 regularizer strength applied to DNN
+    :param l2_reg_embedding: float. L2 regularizer strength applied to embedding vector
+    :param dnn_dropout:  float in [0,1), the probability we will drop out a given DNN coordinate.
+    :param output_activation: Activation function to use in output layer
+    :param sampler_config: negative sample config.
+    :param seed: integer ,to use as random seed.
+    :return: A Keras model instance.
+
+    """
+
+    if len(item_feature_columns) > 1:
+        raise ValueError("Now ComiRec only support 1 item feature like item_id")
+    if interest_extractor.lower() not in ['dr', 'sa']:
+        raise ValueError("Now ComiRec only support dr and sa two interest_extractor")
+    item_feature_column = item_feature_columns[0]
+    item_feature_name = item_feature_column.name
+    item_vocabulary_size = item_feature_columns[0].vocabulary_size
+    item_embedding_dim = item_feature_columns[0].embedding_dim
+    if user_dnn_hidden_units[-1] != item_embedding_dim:
+        user_dnn_hidden_units = tuple(list(user_dnn_hidden_units) + [item_embedding_dim])
+    # item_index = Input(tensor=tf.constant([list(range(item_vocabulary_size))]))
+
+    history_feature_list = [item_feature_name]
+
+    features = build_input_features(user_feature_columns)
+    sparse_feature_columns = list(
+        filter(lambda x: isinstance(x, SparseFeat), user_feature_columns)) if user_feature_columns else []
+    dense_feature_columns = list(
+        filter(lambda x: isinstance(x, DenseFeat), user_feature_columns)) if user_feature_columns else []
+    varlen_sparse_feature_columns = list(
+        filter(lambda x: isinstance(x, VarLenSparseFeat), user_feature_columns)) if user_feature_columns else []
+    history_feature_columns = []
+    sparse_varlen_feature_columns = []
+    history_fc_names = list(map(lambda x: "hist_" + x, history_feature_list))
+    for fc in varlen_sparse_feature_columns:
+        feature_name = fc.name
+        if feature_name in history_fc_names:
+            history_feature_columns.append(fc)
+        else:
+            sparse_varlen_feature_columns.append(fc)
+    seq_max_len = history_feature_columns[0].maxlen
+    inputs_list = list(features.values())
+
+    embedding_matrix_dict = create_embedding_matrix(user_feature_columns + item_feature_columns, l2_reg_embedding,
+                                                    seed=seed, prefix="")
+
+    item_features = build_input_features(item_feature_columns)
+
+    query_emb_list = embedding_lookup(embedding_matrix_dict, item_features, item_feature_columns,
+                                      history_feature_list,
+                                      history_feature_list, to_list=True)
+    keys_emb_list = embedding_lookup(embedding_matrix_dict, features, history_feature_columns, history_fc_names,
+                                     history_fc_names, to_list=True)
+    dnn_input_emb_list = embedding_lookup(embedding_matrix_dict, features, sparse_feature_columns,
+                                          mask_feat_list=history_feature_list, to_list=True)
+    dense_value_list = get_dense_input(features, dense_feature_columns)
+
+    sequence_embed_dict = varlen_embedding_lookup(embedding_matrix_dict, features, sparse_varlen_feature_columns)
+    sequence_embed_list = get_varlen_pooling_list(sequence_embed_dict, features, sparse_varlen_feature_columns,
+                                                  to_list=True)
+
+    dnn_input_emb_list += sequence_embed_list
+
+    # keys_emb = concat_func(keys_emb_list, mask=True)
+    # query_emb = concat_func(query_emb_list, mask=True)
+
+    history_emb = PoolingLayer()(NoMask()(keys_emb_list))  # [None, max_len, emb_dim]
+    target_emb = PoolingLayer()(NoMask()(query_emb_list))
+
+    # target_emb_size = target_emb.get_shape()[-1].value
+    # max_len = history_emb.get_shape()[1].value
+    hist_len = features['hist_len']
+
+    high_capsule = None
+    if interest_extractor.lower() == 'dr':
+        high_capsule = CapsuleLayer(input_units=item_embedding_dim,
+                                    out_units=item_embedding_dim, max_len=seq_max_len,
+                                    k_max=k_max)((history_emb, hist_len))
+    elif interest_extractor.lower() == 'sa':
+        history_emb_add_pos = history_emb
+        if add_pos:
+            position_embedding = PositionEncoding()(history_emb)
+            history_emb_add_pos = add_func([history_emb_add_pos, position_embedding])  # [None, max_len, emb_dim]
+
+        attn = DNN((item_embedding_dim * 4, k_max), activation='tanh', l2_reg=l2_reg_dnn,
+                   dropout_rate=dnn_dropout, use_bn=dnn_use_bn, output_activation=None, seed=seed,
+                   name="user_dnn_attn")(history_emb_add_pos)
+        mask = Lambda(tile_user_his_mask, arguments={'k_max': k_max,
+                                                     'seq_max_len': seq_max_len})(
+            hist_len)  # [None, k_max, max_len]
+
+        high_capsule = Lambda(softmax_Weighted_Sum)((history_emb_add_pos, mask, attn))
+
+    if len(dnn_input_emb_list) > 0 or len(dense_value_list) > 0:
+        user_other_feature = combined_dnn_input(dnn_input_emb_list, dense_value_list)
+        other_feature_tile = Lambda(tile_user_otherfeat, arguments={'k_max': k_max})(user_other_feature)
+        user_deep_input = Concatenate()([NoMask()(other_feature_tile), high_capsule])
+    else:
+        user_deep_input = high_capsule
+
+    user_embeddings = DNN(user_dnn_hidden_units, dnn_activation, l2_reg_dnn,
+                          dnn_dropout, dnn_use_bn, output_activation=output_activation, seed=seed,
+                          name="user_dnn")(
+        user_deep_input)
+
+    item_inputs_list = list(item_features.values())
+
+    item_embedding_matrix = embedding_matrix_dict[item_feature_name]
+
+    item_index = EmbeddingIndex(list(range(item_vocabulary_size)))(item_features[item_feature_name])
+
+    item_embedding_weight = NoMask()(item_embedding_matrix(item_index))
+
+    pooling_item_embedding_weight = PoolingLayer()([item_embedding_weight])
+
+    user_embedding_final = LabelAwareAttention(k_max=k_max, pow_p=p)((user_embeddings, target_emb))
+
+    output = SampledSoftmaxLayer(sampler_config._asdict())(
+        [pooling_item_embedding_weight, user_embedding_final, item_features[item_feature_name]])
+    model = Model(inputs=inputs_list + item_inputs_list, outputs=output)
+
+    model.__setattr__("user_input", inputs_list)
+    model.__setattr__("user_embedding", user_embeddings)
+
+    model.__setattr__("item_input", item_inputs_list)
+    model.__setattr__("item_embedding",
+                      get_item_embedding(pooling_item_embedding_weight, item_features[item_feature_name]))
+
+    return model

deepmatch/models/mind.py

@@ -53,11 +53,9 @@ def MIND(user_feature_columns, item_feature_columns, k_max=2, p=100, dynamic_k=F
 
     :param user_feature_columns: An iterable containing user's features used by  the model.
     :param item_feature_columns: An iterable containing item's features used by  the model.
-    :param num_sampled: int, the number of classes to randomly sample per batch.
     :param k_max: int, the max size of user interest embedding
     :param p: float,the parameter for adjusting the attention distribution in LabelAwareAttention.
     :param dynamic_k: bool, whether or not use dynamic interest number
-    :param dnn_use_bn: bool. Whether use BatchNormalization before activation or not in deep net
     :param user_dnn_hidden_units: list,list of positive integer or empty list, the layer number and units in each layer of user tower
     :param dnn_activation: Activation function to use in deep net
     :param dnn_use_bn: bool. Whether use BatchNormalization before activation or not in deep net
@@ -169,7 +167,7 @@ def MIND(user_feature_columns, item_feature_columns, k_max=2, p=100, dynamic_k=F
         user_embedding_final = LabelAwareAttention(k_max=k_max, pow_p=p)((user_embeddings, target_emb, interest_num))
     else:
         user_embedding_final = LabelAwareAttention(k_max=k_max, pow_p=p)((user_embeddings, target_emb))
-    print("swc")
+
     output = SampledSoftmaxLayer(sampler_config._asdict())(
         [pooling_item_embedding_weight, user_embedding_final, item_features[item_feature_name]])
     model = Model(inputs=inputs_list + item_inputs_list, outputs=output)

deepmatch/models/sdm.py

@@ -30,7 +30,6 @@ def SDM(user_feature_columns, item_feature_columns, history_feature_list, units=
     :param user_feature_columns: An iterable containing user's features used by  the model.
     :param item_feature_columns: An iterable containing item's features used by  the model.
     :param history_feature_list: list,to indicate short and prefer sequence sparse field
-    :param num_sampled: int, the number of classes to randomly sample per batch.
     :param units: int, dimension for each output layer
     :param rnn_layers: int, layer number of rnn
     :param dropout_rate: float in [0,1), the probability we will drop out a given DNN coordinate.

deepmatch/models/youtubednn.py

@@ -23,7 +23,6 @@ def YoutubeDNN(user_feature_columns, item_feature_columns,
 
     :param user_feature_columns: An iterable containing user's features used by  the model.
     :param item_feature_columns: An iterable containing item's features used by  the model.
-    :param num_sampled: int, the number of classes to randomly sample per batch.
     :param user_dnn_hidden_units: list,list of positive integer or empty list, the layer number and units in each layer of user tower
     :param dnn_activation: Activation function to use in deep net
     :param dnn_use_bn: bool. Whether use BatchNormalization before activation or not in deep net

docs/source/Features.md

@@ -96,3 +96,10 @@
 
 [Multi-interest network with dynamic routing for recommendation at Tmall](https://arxiv.org/pdf/1904.08030)
 
+### COMIREC (Controllable Multi-Interest Framework for Recommendation)
+
+[**COMIREC Model API**](./deepmatch.models.comirec.html)
+
+![COMIREC](../pics/comirec.jpg)
+
+[Controllable Multi-Interest Framework for Recommendation](https://arxiv.org/pdf/2005.09347)

docs/source/History.md

@@ -1,4 +1,5 @@
 # History
+- 10/31/2022 : [v0.3.1](https://github.com/shenweichen/DeepMatch/releases/tag/v0.3.1) released.Add `ComiRec` model .
 - 07/04/2022 : [v0.3.0](https://github.com/shenweichen/DeepMatch/releases/tag/v0.3.0) released.Support different negative sampling strategies, including `inbatch`, `uniform`, `frequency`, `adaptive`.
 - 06/17/2022 : [v0.2.1](https://github.com/shenweichen/DeepMatch/releases/tag/v0.2.1) released.Fix some bugs.
 - 10/12/2020 : [v0.2.0](https://github.com/shenweichen/DeepMatch/releases/tag/v0.2.0) released.Support different initializers for different embedding weights and loading pretrained embeddings. 

docs/source/Models.rst

@@ -9,5 +9,5 @@ DeepMatch Models API
    NCF<deepmatch.models.ncf>
    SDM<deepmatch.models.sdm>
    MIND<deepmatch.models.mind>
-
+   COMIREC<deepmatch.models.comirec>
 

docs/source/conf.py

@@ -26,7 +26,7 @@
 # The short X.Y version
 version = ''
 # The full version, including alpha/beta/rc tags
-release = '0.3.0'
+release = '0.3.1'
 
 
 # -- General configuration ---------------------------------------------------

docs/source/deepmatch.models.comirec.rst

@@ -0,0 +1,7 @@
+deepmatch.models.comirec module
+============================
+
+.. automodule:: deepmatch.models.comirec
+    :members:
+    :no-undoc-members:
+    :no-show-inheritance:

docs/source/deepmatch.models.rst

@@ -9,6 +9,7 @@ Submodules
    deepmatch.models.dssm
    deepmatch.models.fm
    deepmatch.models.mind
+   deepmatch.models.comirec
    deepmatch.models.ncf
    deepmatch.models.sdm
    deepmatch.models.youtubednn

docs/source/index.rst

@@ -37,12 +37,12 @@ You can read the latest code at https://github.com/shenweichen/DeepMatch
 News
 -----
 
+10/31/2022 : Add `ComiRec` . `Changelog <https://github.com/shenweichen/DeepMatch/releases/tag/v0.3.1>`_
+
 07/04/2022 : Support different negative sampling strategies, including `inbatch` , `uniform` , `frequency` , `adaptive` . `Changelog <https://github.com/shenweichen/DeepMatch/releases/tag/v0.3.0>`_
 
 06/17/2022 : Fix some bugs. `Changelog <https://github.com/shenweichen/DeepMatch/releases/tag/v0.2.1>`_
 
-10/12/2020 : Support different initializers for different embedding weights and loading pretrained embeddings. `Changelog <https://github.com/shenweichen/DeepMatch/releases/tag/v0.2.0>`_
-
 DisscussionGroup
 -----------------------