Question

我目前正在运行以下代码，以便根据6个参数预测房屋的价格：

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

housing = pd.read_csv('cal_housing_clean.csv')

X = housing.iloc[:,0:6]
y = housing.iloc[:,6:]

from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.3)

from sklearn.preprocessing import MinMaxScaler
scaler = MinMaxScaler()
scaler.fit(X_train)
X_train = pd.DataFrame(data=scaler.transform(X_train),columns = X_train.columns,index=X_train.index)
X_test = pd.DataFrame(data=scaler.transform(X_test),columns = X_test.columns,index=X_test.index)

X_data = tf.placeholder(dtype = "float", shape=[None,6])
y_target = tf.placeholder(dtype = "float", shape=[None,1])

hidden_layer_nodes = 10

w1 = tf.Variable(tf.random_normal(shape=[6,hidden_layer_nodes]))
b1 = tf.Variable(tf.random_normal(shape=[hidden_layer_nodes]))
w2 = tf.Variable(tf.random_normal(shape=[hidden_layer_nodes,1]))
b2 = tf.Variable(tf.random_normal(shape=[1]))

hidden_output = tf.nn.relu(tf.add(tf.matmul(X_data,w1),b1))
y_output = tf.add(tf.matmul(hidden_output,w2),b2)

loss = tf.reduce_mean(tf.square(y_target-y_output))

optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.00001)
train = optimizer.minimize(loss)

init = tf.global_variables_initializer()

steps = 100000

with tf.Session() as sess:

    sess.run(init)

    for i in range(steps):

        sess.run(train, feed_dict={X_data:X_train,y_target:y_train})

        if i%500 == 0:

            print('Currently on step {}'.format(i))

            training_cost = sess.run(loss, feed_dict={X_data:X_test,y_target:y_test})
            print("Training cost=", training_cost/6192)

    training_cost = sess.run(loss, feed_dict={X_data:X_test,y_target:y_test})
    print("Training cost=", training_cost/6192)

我在此认为，由于test_set包含6192行数据，简单地将总损失或错误除以该值就可以解决问题，但不幸的是我达到了以下输出：

Currently on step 0
Training cost= 9190063.95866
Currently on step 500
Training cost= 9062077.85013
Currently on step 1000
Training cost= 8927415.89664
Currently on step 1500
Training cost= 8795428.38243
Currently on step 2000
Training cost= 8666037.25065
Currently on step 2500
Training cost= 8539182.30491
Currently on step 3000
Training cost= 8414841.71576

其中错误将降至约200万，而我希望值接近100或20万。

我的代码中可能存在错误，使得近似值非常糟糕。我也尝试了不同的learning_rates，结果相同。

我还想尝试通过分批发送测试数据来测试模型。我试过这个：

if i%500 == 0:

    rand_ind = np.random.randint(len(X_test),size=8)

    feed = {X_data:X_test[rand_ind],y_target:y_test[rand_ind]}

    loss = tf.reduce_sum(tf.square(y_target-y_output)) / 8

    print(sess.run(loss,feed_dict=feed))

但遗憾的是我总是被告知索引，我选择使用rand_ind＆＃34;而不是索引＆＃34;。

Answer 1

您可以尝试使用tf.train.AdamOptimizer并提高学习率（可能大约为0.1）。这将提高收敛速度。

回归运动中损失惨重！批量输入测试？

1 个答案: