implicit Matrix Factorization - keras

iMF.py

import numpy as np
from numpy import bincount,ravel,log
from scipy.sparse import coo_matrix
import pandas as pd
from keras import backend as K
from keras.layers import *
from keras.models import Model,Sequential

def imf_loss(y_true, y_pred):
    C=1+2*y_true
    y=K.cast(K.cast(y_true, bool),'float32')
    diff = C*K.square(y - y_pred)
    return K.mean(diff, axis=-1)

viewer=Input(shape=(1,),name='resp')
program=Input(shape=(1,),name='prog')
r=Embedding(input_dim=24587,input_length=1,output_dim=16,embeddings_regularizer='l2')(viewer)
p=Embedding(input_dim=6771,input_length=1,output_dim=16,embeddings_regularizer='l2')(program)
r=Flatten()(r)
p=Flatten()(p)
cat=concatenate([r,p])
d=Dense(units=4,activation='sigmoid',kernel_regularizer='l1')(cat)
output=Dense(units=1,activation='sigmoid',kernel_regularizer='l1')(d)
model=Model(inputs=[viewer,program],outputs=[output])

model.summary()
model.compile(optimizer='adam',loss=imf_loss)
model.fit([respondent,program],y_true)