FM算法keras實現

import numpy as np
import pandas as pd
import tensorflow as tf
import keras
import os

import matplotlib.pyplot as plt

from keras.layers import Layer,Dense,Dropout,Input
from keras import Model,activations
from keras.optimizers import Adam
from keras import backend as K
from keras.layers import Layer
from sklearn.datasets import load_breast_cancer

os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ['CUDA_VISIBLE_DEVICES'] = "0"
class FM(Layer):
    def __init__(self, output_dim, latent=10,  activation='relu', **kwargs):
        self.latent = latent
        self.output_dim = output_dim
        self.activation = activations.get(activation)
        super(FM, self).__init__(**kwargs)

    def build(self, input_shape):
        self.b = self.add_weight(name='W0',
                                  shape=(self.output_dim,),
                                  trainable=True,
                                 initializer='zeros')
        self.w = self.add_weight(name='W',
                                 shape=(input_shape[1], self.output_dim),
                                 trainable=True,
                                 initializer='random_uniform')
        self.v= self.add_weight(name='V',
                                 shape=(input_shape[1], self.latent),
                                 trainable=True,
                                initializer='random_uniform')
        super(FM, self).build(input_shape)

    def call(self, inputs, **kwargs):
        x = inputs
        x_square = K.square(x)

        xv = K.square(K.dot(x, self.v))
        xw = K.dot(x, self.w)

        p = 0.5*K.sum(xv-K.dot(x_square, K.square(self.v)), 1)

        rp = K.repeat_elements(K.reshape(p, (-1, 1)), self.output_dim, axis=-1)

        f = xw + rp + self.b

        output = K.reshape(f, (-1, self.output_dim))

        return output

    def compute_output_shape(self, input_shape):
        assert input_shape and len(input_shape)==2
        return input_shape[0],self.output_dim


data = load_breast_cancer()["data"]
target = load_breast_cancer()["target"]

K.clear_session()
print(target)
inputs = Input(shape=(30,))
out = FM(20)(inputs)
out = Dense(15, activation='sigmoid')(out)
out = Dense(1, activation='sigmoid')(out)

model=Model(inputs=inputs, outputs=out)
model.compile(loss='mse',
              optimizer='adam',
              metrics=['acc'])
model.summary()

h=model.fit(data, target, batch_size=1, epochs=10, validation_split=0.2)

#%%

plt.plot(h.history['acc'],label='acc')
plt.plot(h.history['val_acc'],label='val_acc')
plt.xlabel('epoch')
plt.ylabel('acc')

#%%