如何创建 Cifar-10 子集？

Question

I would like to train a deep neural network using fewer training data samples to reduce the time for testing my code. II wanted to know how to subset the Cifar-10 dataset using Keras TensorFlow.I have the following code which is training for Cifar-10 complete dataset.

#load and prepare data
if WhichDataSet == 'CIFAR10':
    (x_train, y_train), (x_test, y_test) = tensorflow.keras.datasets.cifar10.load_data()
else:
    (x_train, y_train), (x_test, y_test) = tensorflow.keras.datasets.cifar100.load_data()
num_classes = np.unique(y_train).shape[0]
K_train = x_train.shape[0]
input_shape = x_train.shape[1:]
x_train = x_train.astype('float32')
x_test = x_test.astype('float32')
y_train = tensorflow.keras.utils.to_categorical(y_train, num_classes)
y_test = tensorflow.keras.utils.to_categorical(y_test, num_classes)

Answer 1

Use the pandas module to create a data frame and sample it accordingly.

import pandas as pd
(train_images1, train_labels), (test_images1, test_labels) = datasets.cifar10.load_data()

# Normalize pixel values to be between 0 and 1
train_images, test_images = train_images1 / 255.0, test_images1 / 255.0

#creating the validation set from the training set
df = pd.DataFrame(list(zip(train_images, train_labels)), columns =['Image', 'label']) 
val = df.sample(frac=0.2)
X_train = np.array([ i for i in list(val['Image'])])
y_train = np.array([ [i[0]] for i in list(val['label'])])

The line val = df.sample(frac=0.2) samples out 0.20 percent of the total data.

You can use val = df.sample(n=5000) if you want a specific number of data records, by setting the n value accordingly.

You can use random_state = 0 if you want the same results every time you run the code. eg:

val = df.sample(n=5000,random_state = 0)

Answer 2

Create susbset based on labels

Create a subset of dataset excluding few labels. For example, to create a new train dataset with only first five class labels you can use below code

subset_x_train = x_train[np.isin(y_train, [0,1,2,3,4]).flatten()]
subset_y_train = y_train[np.isin(y_train, [0,1,2,3,4]).flatten()]

Create subset irrespective of labels

To create a 10% subset of train data you can use below code

# Shuffle first (optional)
idx = np.arange(len(x_train))
np.random.shuffle(idx)

# get first 10% of data
subset_x_train = x_train[:int(.10*len(idx))]
subset_y_train = y_train[:int(.10*len(idx))]

Repeat the same for x_test and y_test to get a subset of test data.