python中MNIST数据中的位数答案

【问题标题】：Number of digits in MNIST data in pythonpython中MNIST数据中的位数
【发布时间】：2022-10-25 19:13:29
【问题描述】：

如何找到训练和测试集中每个数字的样本数？此外，如果我想为训练数据中的每个数字 > 可视化至少 9 个图像。 #在Python中加载MNIST数据集

   no_of_different_labels = 10
   image_pixels = image_size * image_size```

#Loading Training Samples

```train_data = np.loadtxt("mnist_train.csv", delimiter=",")
X_train = train_data[:,1:]
t_train = train_data[:,0]```

#Loading Test Samples

```test_data = np.loadtxt("mnist_test.csv", delimiter=",")` 
X_test = test_data[:,1:]
t_test = test_data[:,0]```


Question: How can I find the number of samples per digit in training and test sets? Also if I want to >visualize at least 9 images for each digit in the training data.

【问题讨论】：

标签： python python-3.x jupyter-notebook mnist anaconda3

【解决方案1】：

我认为您可以使用 Numpy 的 unique 函数，并将结果数字和计数压缩到指示数字计数的字典中：

# y_train is the same as your t_train. It's usual to name label arrays with "y" :-)

# Count elements, and return unique numbers and their counts
unique, counts = np.unique(y_train, return_counts=True)
>>> dict(zip(unique, counts))
{0.0: 5923, 1.0: 6742, 2.0: 5958, 3.0: 6131, 4.0: 5842, 5.0: 5421, 6.0: 5918, 7.0: 6265, 8.0: 5851, 9.0: 5949}

# The same goes for the test set!

此外，如果您想为每个数字可视化一些图像，您可以使用 Python 惊人的 matplotlib 库:-)

import matplotlib.pyplot as plt

# Reshape digits into 28x28 for Matplotlib to properly print them
X_train = X_train.reshape(60000,28,28)

# Create a 2D grid of subfigures
fig, axes = plt.subplots(10, 9, figsize=(15,18))

# i is the digit iterator. j is the samples-per-digit iterator
for i in range(10):
    # Select digits that equal "i"
    digits = X_train[y_train == i]
    for j in range(9):
        # Select axis, and print the digit on it!
        ax = axes[i, j]
        ax.imshow(digits[j], cmap="gray")

# This will give you the visualization :-)
plt.show()

我希望这是有帮助的。干杯!

【讨论】：