from tensorflow.keras import *
from tensorflow.keras.layers import *
from tensorflow.keras.models import *
from tensorflow.keras.callbacks import *
import tensorflow.keras.backend as K
from sklearn.model_selection import train_test_split
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
from sklearn.datasets import load_breast_cancer
from tqdm import tqdm
import pandas as pd
import numpy as np

# FM 特征组合层
class crossLayer(layers.Layer):
    def __init__(self, input_dim, output_dim=10, **kwargs):
        super(crossLayer, self).__init__(**kwargs)
        self.input_dim = input_dim
        self.output_dim = output_dim
        self.kernel = self.add_weight(name='kernel',
                                      shape=(self.input_dim, self.output_dim),
                                      initializer='glorot_uniform',
                                      trainable=True)
    def call(self, x):
        a = K.pow(K.dot(x, self.kernel), 2)
        b = K.dot(K.pow(x, 2), K.pow(self.kernel, 2))
        return 0.5 * K.mean(a - b, 1, keepdims=True)

# 定义FM模型
def FM(feature_dim):
    inputs = Input(shape=(feature_dim,))
    # 一阶特征
    linear = Dense(units=1,
                   kernel_regularizer=regularizers.l2(0.01),
                   bias_regularizer=regularizers.l2(0.01))(inputs)
    # 二阶特征
    cross = crossLayer(feature_dim)(inputs)
    add = Add()([linear, cross])
    pred = Dense(units=1, activation="sigmoid")(add)
    model = Model(inputs=inputs, outputs=pred)
    model.summary()
    model.compile(loss='binary_crossentropy',
                  optimizer=optimizers.Adam(),
                  metrics=['binary_accuracy'])
    return model

# 读取数据
print('loading data...')
X, y = load_breast_cancer(return_X_y = True)

# 划分数据集
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=2021)

# 构建模型
model = FM(X.shape[1])

# 训练模型
model.fit(X_train, y_train, epochs=10, batch_size=128, validation_data=(X_test, y_test))