Question

Python和深度学习的新手。我试图用一些数据建立一个RNN而且我不知道我哪里出错了。

这是我的代码：

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline



raw = pd.read_excel('Online Retail.xlsx',index_col='InvoiceDate')
sales = raw.drop(['InvoiceNo','StockCode','Country','Description'],axis=1)
sales.head()
sales.index = pd.to_datetime(sales.index)
sales.info()

train_set = sales.head(50000)
test_set = sales.tail(41909)

from sklearn.preprocessing import MinMaxScaler
scaler = MinMaxScaler()


training =  np.nan_to_num(train_set)
testing = np.nan_to_num(test_set)

train_scaled = scaler.fit_transform(training)
test_scaled = scaler.fit_transform(testing)


def next_batch(training_data,batch_size,steps):
  rand_start = np.random.randint(0,len(training_data)-steps)
  y_batch = 
  np.array(training_data[rand_start:rand_start+steps+1].reshape(26,steps+1))
  return 
  y_batch[:,:-1].reshape(-1,steps,1),y_batch[:,1:].reshape(-1,steps,1)

  import tensorflow as tf


  num_inputs = 1
  num_time_steps = 10
  num_neurons = 100
  num_outputs = 1
  learning_rate = 0.03
  num_train_iterations = 4000
  batch_size = 1

  X = tf.placeholder(tf.float32,[None,num_time_steps,num_inputs])
  y = tf.placeholder(tf.float32,[None,num_time_steps,num_outputs])

  cell = tf.contrib.rnn.OutputProjectionWrapper(
 tf.contrib.rnn.BasicLSTMCell(num_units=num_neurons,activation=tf.nn.relu),output_size=num_outputs) 

 outputs, states = tf.nn.dynamic_rnn(cell, X, dtype=tf.float32)

 loss = tf.reduce_mean(tf.square(outputs - y)) # MSE
 optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)
 train = optimizer.minimize(loss)

 init = tf.global_variables_initializer()
 saver = tf.train.Saver()
 with tf.Session(config=tf.ConfigProto()) as sess:
     sess.run(init)

     for iteration in range(num_train_iterations):

        X_batch, y_batch = next_batch(train_scaled,batch_size,num_time_steps)
        sess.run(train, feed_dict={X: X_batch, y: y_batch})

        if iteration % 100 == 0:

            mse = loss.eval(feed_dict={X: X_batch, y: y_batch})
            print(iteration, "\tMSE:", mse)

    # Save Model for Later
    saver.save(sess, "./ex_time_series_model")

输出：

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-36-f2f7c66a33df> in <module>()
      4         for iteration in range(num_train_iterations):
      5 
----> 6             X_batch, y_batch = next_batch(train_scaled,batch_size,num_time_steps)
      7             sess.run(train, feed_dict={X: X_batch, y: y_batch})
      8 

<ipython-input-26-f673a469c67d> in next_batch(training_data, batch_size, steps)
      1 def next_batch(training_data,batch_size,steps):
      2     rand_start = np.random.randint(0,len(training_data)-steps)
----> 3     y_batch = np.array(training_data[rand_start:rand_start+steps+1].reshape(26,steps+1))
      4     return y_batch[:,:-1].reshape(-1,steps,1),y_batch[:,1:].reshape(-1,steps,1)

ValueError: cannot reshape array of size 33 into shape (26,11)

In [ ]:

Answer 1

我不确定26号码的来源，但它与您的数据维度不匹配。删除四列后，training_data数组为(50000, 3)，您可以在其中进行(11, 3)次批处理。这个数组显然不能重塑为(26, 11)。

你可能意味着这个（在next_batch函数中）：

y_batch = np.array(training_data[rand_start:rand_start+steps+1].reshape(3,steps+1))

Answer 2

错误表示您尝试将尺寸为33的张量重塑为尺寸为26x11的张量，这是您无法做到的。您应该将大小为286的张量重塑为26x11。

尝试使用next_batch在每个步骤中打印y_batch形状来调试print (y_batch.get_shape())函数，并检查它是否具有286形状。

我没有抓住这一点，为什么你要随机取出每一批？为什么你没有正常读取输入数据？

如果在发布代码时修复缩进量会很好，但很难跟踪。

在时间序列RNN初学者块中重塑阵列

2 个答案: