Tensorflow-“无法以字节形式获取元素”错误

Question

import numpy as np
import tensorflow as tf
from sklearn.model_selection import train_test_split
import pandas as pd



# DATA PREPARE
df = pd.read_csv('housing.csv')
df = df.dropna()
print(df.head)
print(df.describe())
print(df.info())



# NORMALIZATION
from sklearn.preprocessing import MinMaxScaler

scaler = MinMaxScaler()
scaler.fit(df[['housing_median_age', 'total_rooms', 'total_bedrooms', 'population', 'households', 'median_income',
               'median_house_value']])
df_scaled_cols = scaler.transform(df[['housing_median_age', 'total_rooms', 'total_bedrooms',
                                      'population', 'households', 'median_income', 'median_house_value']])
df_scaled_cols = pd.DataFrame(data=df_scaled_cols, columns=['housing_median_age', 'total_rooms', 'total_bedrooms',
                                                            'population', 'households', 'median_income',
                                                            'median_house_value'])

df = pd.concat([df_scaled_cols, df['ocean_proximity']], axis=1)



# DATAFRAME INTO X AND Y -> TRAIN TEST SPLIT
x_data = df[['housing_median_age', 'total_rooms', 'total_bedrooms', 'population', 'households', 'median_income',
             'ocean_proximity']]
y_label = df['median_house_value']

X_train, X_test, y_train, y_test = train_test_split(x_data, y_label, test_size=0.3)



# FEATURE COLUMNS FROM DATA

m_age = tf.feature_column.numeric_column('housing_median_age')
rooms = tf.feature_column.numeric_column('total_rooms')
bedrooms = tf.feature_column.numeric_column('total_bedrooms')
population = tf.feature_column.numeric_column('population')
households = tf.feature_column.numeric_column('households')
income = tf.feature_column.numeric_column('median_income')
ocean = tf.feature_column.categorical_column_with_hash_bucket('ocean_proximity', hash_bucket_size=10)
embedded_ocean = tf.feature_column.embedding_column(ocean, dimension=4)

feat_cols = [m_age, rooms, bedrooms, population, households, income, embedded_ocean]



# 3 INPUT FUNCTIONS

train_input_func = tf.estimator.inputs.pandas_input_fn(x=X_train, y=y_train, batch_size=10, num_epochs=1000,
                                                       shuffle=True)
test_input_func = tf.estimator.inputs.pandas_input_fn(x=X_test, y=y_test, batch_size=10, num_epochs=1, shuffle=False)
predict_input_func = tf.estimator.inputs.pandas_input_fn(x=X_test, batch_size=10, num_epochs=1, shuffle=False)



# DNN_Reg MODEL

dnn_model = tf.estimator.DNNRegressor(hidden_units=[10,10,10], feature_columns=feat_cols)
dnn_model.train(input_fn=train_input_func, steps=1000)

Answer 1

问题是正常化。

代替sklearn方法，我做了以下操作：

df[['housing_median_age', 'total_rooms', 'total_bedrooms', 'population', 'households', 'median_income',
    'median_house_value']] = df[['housing_median_age', 'total_rooms', 'total_bedrooms', 'population', 'households', 'median_income',
               'median_house_value']].apply(lambda x: (x-x.min()) / (x.max()-x.min()))

所以，最后，我做了和sklearn一样的事情，只是手动的--用lambda。