Question

在tensorflow MNIST tutorial中，mnist.train.next_batch(100)功能非常方便。我现在正试图自己实现一个简单的分类。我的训练数据是一个numpy数组。我如何为自己的数据实现类似的功能，以便为我提供下一批产品？

sess = tf.InteractiveSession()
tf.global_variables_initializer().run()
Xtr, Ytr = loadData()
for it in range(1000):
    batch_x = Xtr.next_batch(100)
    batch_y = Ytr.next_batch(100)

Answer 1

您发布的链接说：＆＃34;我们收到了＃34;批次＆＃34;来自我们的训练集＆＃34; 的100个随机数据点。在我的示例中，我使用了一个全局函数（不是您示例中的方法），因此语法会有所不同。

在我的函数中，您需要传递所需的样本数和数据数组。

这是正确的代码，可确保样本具有正确的标签：

import numpy as np

def next_batch(num, data, labels):
    '''
    Return a total of `num` random samples and labels. 
    '''
    idx = np.arange(0 , len(data))
    np.random.shuffle(idx)
    idx = idx[:num]
    data_shuffle = [data[ i] for i in idx]
    labels_shuffle = [labels[ i] for i in idx]

    return np.asarray(data_shuffle), np.asarray(labels_shuffle)

Xtr, Ytr = np.arange(0, 10), np.arange(0, 100).reshape(10, 10)
print(Xtr)
print(Ytr)

Xtr, Ytr = next_batch(5, Xtr, Ytr)
print('\n5 random samples')
print(Xtr)
print(Ytr)

演示运行：

[0 1 2 3 4 5 6 7 8 9]
[[ 0  1  2  3  4  5  6  7  8  9]
 [10 11 12 13 14 15 16 17 18 19]
 [20 21 22 23 24 25 26 27 28 29]
 [30 31 32 33 34 35 36 37 38 39]
 [40 41 42 43 44 45 46 47 48 49]
 [50 51 52 53 54 55 56 57 58 59]
 [60 61 62 63 64 65 66 67 68 69]
 [70 71 72 73 74 75 76 77 78 79]
 [80 81 82 83 84 85 86 87 88 89]
 [90 91 92 93 94 95 96 97 98 99]]

5 random samples
[9 1 5 6 7]
[[90 91 92 93 94 95 96 97 98 99]
 [10 11 12 13 14 15 16 17 18 19]
 [50 51 52 53 54 55 56 57 58 59]
 [60 61 62 63 64 65 66 67 68 69]
 [70 71 72 73 74 75 76 77 78 79]]

Answer 2

为了对每个小批量进行洗牌和取样，还应考虑是否在当前时期内选择了样本。这是一个使用上述答案中的数据的实现。

import numpy as np 

class Dataset:

def __init__(self,data):
    self._index_in_epoch = 0
    self._epochs_completed = 0
    self._data = data
    self._num_examples = data.shape[0]
    pass


@property
def data(self):
    return self._data

def next_batch(self,batch_size,shuffle = True):
    start = self._index_in_epoch
    if start == 0 and self._epochs_completed == 0:
        idx = np.arange(0, self._num_examples)  # get all possible indexes
        np.random.shuffle(idx)  # shuffle indexe
        self._data = self.data[idx]  # get list of `num` random samples

    # go to the next batch
    if start + batch_size > self._num_examples:
        self._epochs_completed += 1
        rest_num_examples = self._num_examples - start
        data_rest_part = self.data[start:self._num_examples]
        idx0 = np.arange(0, self._num_examples)  # get all possible indexes
        np.random.shuffle(idx0)  # shuffle indexes
        self._data = self.data[idx0]  # get list of `num` random samples

        start = 0
        self._index_in_epoch = batch_size - rest_num_examples #avoid the case where the #sample != integar times of batch_size
        end =  self._index_in_epoch  
        data_new_part =  self._data[start:end]  
        return np.concatenate((data_rest_part, data_new_part), axis=0)
    else:
        self._index_in_epoch += batch_size
        end = self._index_in_epoch
        return self._data[start:end]

dataset = Dataset(np.arange(0, 10))
for i in range(10):
    print(dataset.next_batch(5))

输出是：

[2 8 6 3 4]
[1 5 9 0 7]
[1 7 3 0 8]
[2 6 5 9 4]
[1 0 4 8 3]
[7 6 2 9 5]
[9 5 4 6 2]
[0 1 8 7 3]
[9 7 8 1 6]
[3 5 2 4 0]

第一个和第二个（第3个和第4个......）迷你批次对应一个整个时代..

Answer 3

我使用Anaconda和Jupyter。在Jupyter，如果你运行?mnist，你会得到： File: c:\programdata\anaconda3\lib\site-packages\tensorflow\contrib\learn\python\learn\datasets\base.py Docstring: Datasets(train, validation, test)

在文件夹datesets中，您会发现mnist.py包含所有方法，包括next_batch。

Answer 4

上面标出的答案是我通过该算法尝试的算法，但没有得到结果，所以我在kaggle上进行搜索，发现确实很棒的算法效果很好。最好的结果试试这个。 在下面的算法中，“全局变量”采用您在上面声明的输入中读取数据集。**

epochs_completed = 0
index_in_epoch = 0
num_examples = X_train.shape[0]
    # for splitting out batches of data
def next_batch(batch_size):

    global X_train
    global y_train
    global index_in_epoch
    global epochs_completed

    start = index_in_epoch
    index_in_epoch += batch_size

    # when all trainig data have been already used, it is reorder randomly    
    if index_in_epoch > num_examples:
        # finished epoch
        epochs_completed += 1
        # shuffle the data
        perm = np.arange(num_examples)
        np.random.shuffle(perm)
        X_train = X_train[perm]
        y_train = y_train[perm]
        # start next epoch
        start = 0
        index_in_epoch = batch_size
        assert batch_size <= num_examples
    end = index_in_epoch
    return X_train[start:end], y_train[start:end]

Answer 5

如果您不希望在tensorflow会话运行中出现形状不匹配错误然后使用以下函数代替上面第一个解决方案（https://stackoverflow.com/a/40995666/7748451） -

中提供的函数

def next_batch(num, data, labels):

    '''
    Return a total of `num` random samples and labels. 
    '''
    idx = np.arange(0 , len(data))
    np.random.shuffle(idx)
    idx = idx[:num]
    data_shuffle = data[idx]
    labels_shuffle = labels[idx]
    labels_shuffle = np.asarray(labels_shuffle.values.reshape(len(labels_shuffle), 1))

    return data_shuffle, labels_shuffle

Answer 6

另一个实现方式：

from typing import Tuple
import numpy as np

class BatchMaker(object):
    def __init__(self, feat: np.array, lab: np.array) -> None:
        if len(feat) != len(lab):
            raise ValueError("Expected feat and lab to have the same number of samples")
        self.feat = feat
        self.lab = lab
        self.indexes = np.arange(len(feat))
        np.random.shuffle(self.indexes)
        self.pos = 0

    # "BatchMaker, BatchMaker, make me a batch..."
    def next_batch(self, batch_size: int) -> Tuple[np.array, np.array]:
        if self.pos + batch_size > len(self.feat):
            np.random.shuffle(self.indexes)
            self.pos = 0
        batch_indexes = self.indexes[self.pos: self.pos + batch_size]
        self.pos += batch_size
        return self.feat[batch_indexes], self.lab[batch_indexes]

如何为自己的数据实现tensorflow的next_batch

6 个答案: