├── LICENSE
├── README.md
└── wide_deep_keras.py


/LICENSE:
--------------------------------------------------------------------------------
 1 | MIT License
 2 | 
 3 | Copyright (c) 2017 Jonathan Rahn
 4 | 
 5 | Permission is hereby granted, free of charge, to any person obtaining a copy
 6 | of this software and associated documentation files (the "Software"), to deal
 7 | in the Software without restriction, including without limitation the rights
 8 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 9 | copies of the Software, and to permit persons to whom the Software is
10 | furnished to do so, subject to the following conditions:
11 | 
12 | The above copyright notice and this permission notice shall be included in all
13 | copies or substantial portions of the Software.
14 | 
15 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21 | SOFTWARE.
22 | 


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
1 | # keras-wide-n-deep
2 | Reimplementation of [Google's Wide & Deep Network](https://arxiv.org/abs/1606.07792) in Keras
3 | 
4 | Based on a [TF Tutorial](https://www.tensorflow.org/tutorials/wide_and_deep/) and [Liu Sida's Blog Post](https://liusida.github.io/2016/10/31/translate-from-tf-2-keras/).
5 | 
6 | work in progress
7 | 


--------------------------------------------------------------------------------
/wide_deep_keras.py:
--------------------------------------------------------------------------------
 1 | import pandas as pd
 2 | from keras.models import Sequential
 3 | from keras.layers import Dense, Merge
 4 | from sklearn.preprocessing import MinMaxScaler
 5 | 
 6 | COLUMNS = [
 7 |     "age", "workclass", "fnlwgt", "education", "education_num", "marital_status", 
 8 |     "occupation", "relationship", "race", "gender", "capital_gain", "capital_loss", 
 9 |     "hours_per_week", "native_country", "income_bracket"
10 | ]
11 | 
12 | LABEL_COLUMN = "label"
13 | 
14 | CATEGORICAL_COLUMNS = [
15 |     "workclass", "education", "marital_status", "occupation", "relationship", 
16 |     "race", "gender", "native_country"
17 | ]
18 | 
19 | CONTINUOUS_COLUMNS = [
20 |     "age", "education_num", "capital_gain", "capital_loss", "hours_per_week"
21 | ]
22 | 
23 | def load(filename):
24 |     with open(filename, 'r') as f:
25 |         skiprows = 1 if 'test' in filename else 0
26 |         df = pd.read_csv(
27 |             f, names=COLUMNS, skipinitialspace=True, skiprows=skiprows, engine='python'
28 |         )
29 |         df = df.dropna(how='any', axis=0)
30 |     return df
31 | 
32 | def preprocess(df):
33 |     df[LABEL_COLUMN] = df['income_bracket'].apply(lambda x: ">50K" in x).astype(int)
34 |     df.pop("income_bracket")
35 |     y = df[LABEL_COLUMN].values
36 |     df.pop(LABEL_COLUMN)
37 |     
38 |     df = pd.get_dummies(df, columns=[x for x in CATEGORICAL_COLUMNS])
39 | 
40 |     # TODO: select features for wide & deep parts
41 |     
42 |     # TODO: transformations (cross-products)
43 |     # from sklearn.preprocessing import PolynomialFeatures
44 |     # X = PolynomialFeatures(degree=2, interaction_only=True, include_bias=False).fit_transform(X)
45 |     
46 |     df = pd.DataFrame(MinMaxScaler().fit_transform(df), columns=df.columns)
47 | 
48 |     X = df.values
49 |     return X, y
50 | 
51 | def main():
52 |     df_train = load('adult.data')
53 |     df_test = load('adult.test')
54 |     df = pd.concat([df_train, df_test])
55 |     train_len = len(df_train)
56 |     
57 |     X, y = preprocess(df)
58 |     X_train = X[:train_len]
59 |     y_train = y[:train_len]
60 |     X_test = X[train_len:]
61 |     y_test = y[train_len:]
62 |     
63 |     wide = Sequential()
64 |     wide.add(Dense(1, input_dim=X_train.shape[1]))
65 |     
66 |     deep = Sequential()
67 |     # TODO: add embedding
68 |     deep.add(Dense(input_dim=X_train.shape[1], output_dim=100, activation='relu'))
69 |     deep.add(Dense(100, activation='relu'))
70 |     deep.add(Dense(50, activation='relu'))
71 |     deep.add(Dense(1, activation='sigmoid'))
72 |     
73 |     model = Sequential()
74 |     model.add(Merge([wide, deep], mode='concat', concat_axis=1))
75 |     model.add(Dense(1, activation='sigmoid'))
76 |     
77 |     model.compile(
78 |         optimizer='rmsprop',
79 |         loss='binary_crossentropy',
80 |         metrics=['accuracy']
81 |     )
82 |     
83 |     model.fit([X_train, X_train], y_train, nb_epoch=10, batch_size=32)
84 |     
85 |     loss, accuracy = model.evaluate([X_test, X_test], y_test)
86 |     print('\n', 'test accuracy:', accuracy)
87 |     
88 | if __name__ == '__main__':
89 |     main()
90 | 


--------------------------------------------------------------------------------