feat:add FGCNN&FGCNNLayer

shenweichen · Apr 27, 2019 · ab7cdd8 · ab7cdd8
1 parent 7d5bfd9
commit ab7cdd8
Show file tree

Hide file tree

Showing 20 changed files with 311 additions and 42 deletions.
diff --git a/README.md b/README.md
@@ -36,7 +36,8 @@ Let's [**Get Started!**](https://deepctr-doc.readthedocs.io/en/latest/Quick-Star
 |         Deep Interest Network          | [KDD 2018][Deep Interest Network for Click-Through Rate Prediction](https://arxiv.org/pdf/1706.06978.pdf)                                                       |
 |    Deep Interest Evolution Network     | [AAAI 2019][Deep Interest Evolution Network for Click-Through Rate Prediction](https://arxiv.org/pdf/1809.03672.pdf)                                           |
 |                AutoInt                 | [arxiv 2018][AutoInt: Automatic Feature Interaction Learning via Self-Attentive Neural Networks](https://arxiv.org/abs/1810.11921)                              |
-|                  NFFM                  | [arxiv 2019][Field-aware Neural Factorization Machine for Click-Through Rate Prediction ](https://arxiv.org/pdf/1902.09096.pdf)(The original NFFM was first used by Yi Yang(yangyi868@gmail.com) in TSA competition in 2017.)                                 |
+|                  NFFM                  | [arxiv 2019][Field-aware Neural Factorization Machine for Click-Through Rate Prediction ](https://arxiv.org/pdf/1902.09096.pdf) (The original NFFM was first used by Yi Yang(yangyi868@gmail.com) in TSA competition in 2017.)                                 |
+|                  FGCNN                  | [WWW 2019][Feature Generation by Convolutional Neural Network for Click-Through Rate Prediction ](https://arxiv.org/pdf/1904.04447))   
 
 
 
diff --git a/deepctr/__init__.py b/deepctr/__init__.py
@@ -2,5 +2,5 @@
 from . import models
 from .utils import check_version, SingleFeat, VarLenFeat
 
-__version__ = '0.3.3'
+__version__ = '0.3.4'
 check_version(__version__)
diff --git a/deepctr/input_embedding.py b/deepctr/input_embedding.py
@@ -155,10 +155,10 @@ def get_inputs_list(inputs):
 
 def get_inputs_embedding(feature_dim_dict, embedding_size, l2_reg_embedding, l2_reg_linear, init_std, seed,
                          sparse_input_dict, dense_input_dict, sequence_input_dict, sequence_input_len_dict,
-                         sequence_max_len_dict, sequence_pooling_dict, include_linear):
+                         sequence_max_len_dict, sequence_pooling_dict, include_linear,prefix=""):
 
     deep_sparse_emb_dict = create_embedding_dict(
-        feature_dim_dict, embedding_size, init_std, seed, l2_reg_embedding)
+        feature_dim_dict, embedding_size, init_std, seed, l2_reg_embedding,prefix=prefix+'sparse')
 
     deep_emb_list = get_embedding_vec_list(
         deep_sparse_emb_dict, sparse_input_dict)
@@ -171,7 +171,7 @@ def get_inputs_embedding(feature_dim_dict, embedding_size, l2_reg_embedding, l2_
 
     if include_linear:
         linear_sparse_emb_dict = create_embedding_dict(
-            feature_dim_dict, 1, init_std, seed, l2_reg_linear, 'linear')
+            feature_dim_dict, 1, init_std, seed, l2_reg_linear,prefix+ 'linear')
         linear_emb_list = get_embedding_vec_list(
             linear_sparse_emb_dict, sparse_input_dict)
         linear_emb_list = merge_sequence_input(linear_sparse_emb_dict, linear_emb_list, sequence_input_dict,

diff --git a/deepctr/layers/__init__.py b/deepctr/layers/__init__.py
@@ -4,7 +4,7 @@
 from .core import MLP, LocalActivationUnit, PredictionLayer
 from .interaction import (CIN, FM, AFMLayer, BiInteractionPooling, CrossNet,
                           InnerProductLayer, InteractingLayer,
-                          OutterProductLayer)
+                          OutterProductLayer,FGCNNLayer)
 from .normalization import LayerNormalization
 from .sequence import (AttentionSequencePoolingLayer, BiasEncoding, BiLSTM,
                        KMaxPooling, Position_Embedding, SequencePoolingLayer,
@@ -31,4 +31,5 @@
                   'Transformer': Transformer,
                   'NoMask': NoMask,
                   'BiasEncoding': BiasEncoding,
-                  'KMaxPooling': KMaxPooling}
+                  'KMaxPooling': KMaxPooling,
+                  'FGCNNLayer':FGCNNLayer}
diff --git a/deepctr/layers/interaction.py b/deepctr/layers/interaction.py
@@ -16,6 +16,7 @@
 from tensorflow.python.keras.regularizers import l2
 
 from .activation import activation_fun
+from .utils import concat_fun
 
 
 class AFMLayer(Layer):
@@ -113,7 +114,7 @@ def call(self, inputs, **kwargs):
         self.normalized_att_score = tf.nn.softmax(tf.tensordot(
             attention_temp, self.projection_h, axes=(-1, 0)), dim=1)
         attention_output = tf.reduce_sum(
-            self.normalized_att_score*bi_interaction, axis=1)
+            self.normalized_att_score * bi_interaction, axis=1)
 
         attention_output = tf.nn.dropout(
             attention_output, self.keep_prob, seed=1024)
@@ -130,7 +131,7 @@ def compute_output_shape(self, input_shape):
                              'on a list of inputs.')
         return (None, 1)
 
-    def get_config(self,):
+    def get_config(self, ):
         config = {'attention_factor': self.attention_factor,
                   'l2_reg_w': self.l2_reg_w, 'keep_prob': self.keep_prob, 'seed': self.seed}
         base_config = super(AFMLayer, self).get_config()
@@ -175,7 +176,7 @@ def call(self, inputs, **kwargs):
             concated_embeds_value, axis=1, keep_dims=True))
         sum_of_square = tf.reduce_sum(
             concated_embeds_value * concated_embeds_value, axis=1, keep_dims=True)
-        cross_term = 0.5*(square_of_sum - sum_of_square)
+        cross_term = 0.5 * (square_of_sum - sum_of_square)
 
         return cross_term
 
@@ -206,7 +207,7 @@ class CIN(Layer):
         - [Lian J, Zhou X, Zhang F, et al. xDeepFM: Combining Explicit and Implicit Feature Interactions for Recommender Systems[J]. arXiv preprint arXiv:1803.05170, 2018.] (https://arxiv.org/pdf/1803.05170.pdf)
     """
 
-    def __init__(self, layer_size=(128, 128), activation='relu',split_half=True,  l2_reg=1e-5,seed=1024, **kwargs):
+    def __init__(self, layer_size=(128, 128), activation='relu', split_half=True, l2_reg=1e-5, seed=1024, **kwargs):
         if len(layer_size) == 0:
             raise ValueError(
                 "layer_size must be a list(tuple) of length greater than 1")
@@ -230,7 +231,8 @@ def build(self, input_shape):
             self.filters.append(self.add_weight(name='filter' + str(i),
                                                 shape=[1, self.field_nums[-1]
                                                        * self.field_nums[0], size],
-                                                dtype=tf.float32, initializer=glorot_uniform(seed=self.seed + i), regularizer=l2(self.l2_reg)))
+                                                dtype=tf.float32, initializer=glorot_uniform(seed=self.seed + i),
+                                                regularizer=l2(self.l2_reg)))
 
             self.bias.append(self.add_weight(name='bias' + str(i), shape=[size], dtype=tf.float32,
                                              initializer=tf.keras.initializers.Zeros()))
@@ -346,13 +348,13 @@ def build(self, input_shape):
                 "Unexpected inputs dimensions %d, expect to be 2 dimensions" % (len(input_shape),))
 
         dim = input_shape[-1].value
-        self.kernels = [self.add_weight(name='kernel'+str(i),
+        self.kernels = [self.add_weight(name='kernel' + str(i),
                                         shape=(dim, 1),
                                         initializer=glorot_normal(
                                             seed=self.seed),
                                         regularizer=l2(self.l2_reg),
                                         trainable=True) for i in range(self.layer_num)]
-        self.bias = [self.add_weight(name='bias'+str(i),
+        self.bias = [self.add_weight(name='bias' + str(i),
                                      shape=(dim, 1),
                                      initializer=Zeros(),
                                      trainable=True) for i in range(self.layer_num)]
@@ -373,7 +375,7 @@ def call(self, inputs, **kwargs):
         x_l = tf.squeeze(x_l, axis=2)
         return x_l
 
-    def get_config(self,):
+    def get_config(self, ):
 
         config = {'layer_num': self.layer_num,
                   'l2_reg': self.l2_reg, 'seed': self.seed}
@@ -436,7 +438,7 @@ class InnerProductLayer(Layer):
     product or inner product between feature vectors.
 
       Input shape
-        - A list of N 3D tensor with shape: ``(batch_size,1,embedding_size)``.
+        - a list of 3D tensor with shape: ``(batch_size,1,embedding_size)``.
 
       Output shape
         - 3D tensor with shape: ``(batch_size, N*(N-1)/2 ,1)`` if use reduce_sum. or 3D tensor with shape: ``(batch_size, N*(N-1)/2, embedding_size )`` if not use reduce_sum.
@@ -492,7 +494,9 @@ def call(self, inputs, **kwargs):
                 col.append(j)
         p = tf.concat([embed_list[idx]
                        for idx in row], axis=1)  # batch num_pairs k
-        q = tf.concat([embed_list[idx] for idx in col], axis=1)
+        q = tf.concat([embed_list[idx]
+                       for idx in col], axis=1)
+
         inner_product = p * q
         if self.reduce_sum:
             inner_product = tf.reduce_sum(
@@ -509,7 +513,7 @@ def compute_output_shape(self, input_shape):
         else:
             return (input_shape[0], num_pairs, embed_size)
 
-    def get_config(self,):
+    def get_config(self, ):
         config = {'reduce_sum': self.reduce_sum, }
         base_config = super(InnerProductLayer, self).get_config()
         return dict(list(base_config.items()) + list(config.items()))
@@ -549,14 +553,18 @@ def build(self, input_shape):
             raise ValueError(
                 "Unexpected inputs dimensions %d, expect to be 3 dimensions" % (len(input_shape)))
         embedding_size = input_shape[-1].value
-        self.W_Query = self.add_weight(name='query', shape=[embedding_size, self.att_embedding_size * self.head_num], dtype=tf.float32,
+        self.W_Query = self.add_weight(name='query', shape=[embedding_size, self.att_embedding_size * self.head_num],
+                                       dtype=tf.float32,
                                        initializer=tf.keras.initializers.TruncatedNormal(seed=self.seed))
-        self.W_key = self.add_weight(name='key', shape=[embedding_size, self.att_embedding_size * self.head_num], dtype=tf.float32,
-                                     initializer=tf.keras.initializers.TruncatedNormal(seed=self.seed+1))
-        self.W_Value = self.add_weight(name='value', shape=[embedding_size, self.att_embedding_size * self.head_num], dtype=tf.float32,
-                                       initializer=tf.keras.initializers.TruncatedNormal(seed=self.seed+2))
+        self.W_key = self.add_weight(name='key', shape=[embedding_size, self.att_embedding_size * self.head_num],
+                                     dtype=tf.float32,
+                                     initializer=tf.keras.initializers.TruncatedNormal(seed=self.seed + 1))
+        self.W_Value = self.add_weight(name='value', shape=[embedding_size, self.att_embedding_size * self.head_num],
+                                       dtype=tf.float32,
+                                       initializer=tf.keras.initializers.TruncatedNormal(seed=self.seed + 2))
         if self.use_res:
-            self.W_Res = self.add_weight(name='res', shape=[embedding_size, self.att_embedding_size * self.head_num], dtype=tf.float32,
+            self.W_Res = self.add_weight(name='res', shape=[embedding_size, self.att_embedding_size * self.head_num],
+                                         dtype=tf.float32,
                                          initializer=tf.keras.initializers.TruncatedNormal(seed=self.seed))
 
         # Be sure to call this somewhere!
@@ -656,10 +664,12 @@ def build(self, input_shape):
         embed_size = input_shape[-1].value
         if self.kernel_type == 'mat':
 
-            self.kernel = self.add_weight(shape=(embed_size, num_pairs, embed_size), initializer=glorot_uniform(seed=self.seed),
+            self.kernel = self.add_weight(shape=(embed_size, num_pairs, embed_size),
+                                          initializer=glorot_uniform(seed=self.seed),
                                           name='kernel')
         elif self.kernel_type == 'vec':
-            self.kernel = self.add_weight(shape=(num_pairs, embed_size,), initializer=glorot_uniform(self.seed), name='kernel'
+            self.kernel = self.add_weight(shape=(num_pairs, embed_size,), initializer=glorot_uniform(self.seed),
+                                          name='kernel'
                                           )
         elif self.kernel_type == 'num':
             self.kernel = self.add_weight(
@@ -737,7 +747,88 @@ def compute_output_shape(self, input_shape):
         num_pairs = int(num_inputs * (num_inputs - 1) / 2)
         return (None, num_pairs)
 
-    def get_config(self,):
+    def get_config(self, ):
         config = {'kernel_type': self.kernel_type, 'seed': self.seed}
         base_config = super(OutterProductLayer, self).get_config()
         return dict(list(base_config.items()) + list(config.items()))
+
+
+class FGCNNLayer(Layer):
+    """Feature Generation Layer used in FGCNN,including Convolution,MaxPooling and Recombination.
+
+      Input shape
+        - A 3D tensor with shape:``(batch_size,field_size,embedding_size)``.
+
+      Output shape
+        - 3D tensor with shape: ``(batch_size,new_feture_num,embedding_size)``.
+
+      References
+        - [Liu B, Tang R, Chen Y, et al. Feature Generation by Convolutional Neural Network for Click-Through Rate Prediction[J]. arXiv preprint arXiv:1904.04447, 2019.](https://arxiv.org/pdf/1904.04447)
+
+    """
+
+    def __init__(self, filters=(14, 16,), kernel_width=(7, 7,), new_maps=(3, 3,), pooling_width=(2, 2),
+                 **kwargs):
+        if not (len(filters) == len(kernel_width) == len(new_maps) == len(pooling_width)):
+            raise ValueError("length of argument must be equal")
+        self.filters = filters
+        self.kernel_width = kernel_width
+        self.new_maps = new_maps
+        self.pooling_width = pooling_width
+
+        super(FGCNNLayer, self).__init__(**kwargs)
+
+    def build(self, input_shape):
+
+        if len(input_shape) != 3:
+            raise ValueError(
+                "Unexpected inputs dimensions %d, expect to be 3 dimensions" % (len(input_shape)))
+
+        super(FGCNNLayer, self).build(
+            input_shape)  # Be sure to call this somewhere!
+
+    def call(self, inputs, **kwargs):
+
+        if K.ndim(inputs) != 3:
+            raise ValueError(
+                "Unexpected inputs dimensions %d, expect to be 3 dimensions" % (K.ndim(inputs)))
+
+        embedding_size = inputs.shape[-1].value
+        pooling_result = tf.keras.layers.Lambda(lambda x: tf.expand_dims(x, axis=3))(inputs)
+
+        new_feature_list = []
+
+        for i in range(1, len(self.filters) + 1):
+            filters = self.filters[i - 1]
+            width = self.kernel_width[i - 1]
+            new_filters = self.new_maps[i - 1]
+            pooling_width = self.pooling_width[i - 1]
+            conv_result = tf.keras.layers.Conv2D(filters=filters, kernel_size=(width, 1), strides=(1, 1),
+                                                 padding='same',
+                                                 activation='tanh', use_bias=True, )(pooling_result)
+            pooling_result = tf.keras.layers.MaxPooling2D(pool_size=(pooling_width, 1))(conv_result)
+            flatten_result = tf.keras.layers.Flatten()(pooling_result)
+            new_result = tf.keras.layers.Dense(pooling_result.shape[1].value * embedding_size * new_filters,
+                                               activation='tanh', use_bias=True)(flatten_result)
+            new_feature_list.append(
+                tf.keras.layers.Reshape((pooling_result.shape[1].value * new_filters, embedding_size))(new_result))
+        new_features = concat_fun(new_feature_list, axis=1)
+        return new_features
+
+    def compute_output_shape(self, input_shape):
+
+        new_features_num = 0
+        features_num = input_shape[1]
+
+        for i in range(0, len(self.kernel_width)):
+            pooled_features_num = features_num // self.pooling_width[i]
+            new_features_num += self.new_maps[i] * pooled_features_num
+            features_num = pooled_features_num
+
+        return (None, new_features_num, input_shape[-1])
+
+    def get_config(self, ):
+        config = {'kernel_width': self.kernel_width, 'filters': self.filters, 'new_maps': self.new_maps,
+                  'pooling_width': self.pooling_width}
+        base_config = super(FGCNNLayer, self).get_config()
+        return dict(list(base_config.items()) + list(config.items()))
diff --git a/deepctr/models/__init__.py b/deepctr/models/__init__.py
@@ -12,6 +12,7 @@
 from .pnn import PNN
 from .wdl import WDL
 from .xdeepfm import xDeepFM
+from .fgcnn import FGCNN
 
 __all__ = ["AFM", "CCPM","DCN", "MLR",  "DeepFM",
-           "MLR", "NFM", "DIN", "DIEN", "FNN", "PNN", "WDL", "xDeepFM", "AutoInt", "NFFM"]
+           "MLR", "NFM", "DIN", "DIEN", "FNN", "PNN", "WDL", "xDeepFM", "AutoInt", "NFFM","FGCNN"]
diff --git a/deepctr/models/ccpm.py b/deepctr/models/ccpm.py
@@ -39,6 +39,8 @@ def CCPM(feature_dim_dict, embedding_size=8, conv_kernel_width=(6, 5), conv_filt
     """
 
     check_feature_config_dict(feature_dim_dict)
+    if len(conv_kernel_width)!=len(conv_filters):
+        raise ValueError("conv_kernel_width must have same element with conv_filters")
 
     deep_emb_list, linear_logit, inputs_list = preprocess_input_embedding(feature_dim_dict, embedding_size,
                                                                           l2_reg_embedding, l2_reg_linear, init_std,