embedding_rec.py

def EmbeddingRec(EMBEDDING_SIZE, NUM_MOVIES, NUM_USERS, ROW_COUNT):
    movie_input = keras.Input(shape=(1,), name='movie_id')

    movie_emb = layers.Embedding(output_dim=EMBEDDING_SIZE, input_dim=NUM_MOVIES, input_length=ROW_COUNT, name='movie_emb')(movie_input)
    movie_vec = layers.Flatten(name='FlattenMovie')(movie_emb)

    movie_model = keras.Model(inputs=movie_input, outputs=movie_vec)
    
    user_input = keras.Input(shape=(1,), name='user_id')

    user_emb = layers.Embedding(output_dim=EMBEDDING_SIZE, input_dim=NUM_USERS, input_length=ROW_COUNT, name='user_emb')(user_input)
    user_vec = layers.Flatten(name='FlattenUser')(user_emb)

    user_model = keras.Model(inputs=user_input, outputs=user_vec)
    
    merged = layers.Dot(name = 'dot_product', normalize = True, axes = 2)([movie_emb, user_emb])
    merged_dropout = layers.Dropout(0.2)(merged)
    
    
    dense_1 = layers.Dense(70,name='FullyConnected-1')(merged)
    dropout_1 = layers.Dropout(0.2,name='Dropout_1')(dense_1)

    dense_2 = layers.Dense(50,name='FullyConnected-2')(dropout_1)
    dropout_2 = layers.Dropout(0.2,name='Dropout_2')(dense_2)

    dense_3 = keras.layers.Dense(20,name='FullyConnected-3')(dropout_2)
    dropout_3 = keras.layers.Dropout(0.2,name='Dropout_3')(dense_3)

    dense_4 = keras.layers.Dense(10,name='FullyConnected-4', activation='relu')(dropout_3)

    result = layers.Dense(1, name='result', activation="relu") (dense_4)

    adam = keras.optimizers.Adam(lr=0.001)
    model = keras.Model([movie_input, user_input], result)
    model.compile(optimizer=adam,loss= 'mean_absolute_error')
    return model, movie_model, user_model