When learning machine learning, first and foremost task is to prepare a common set of data to facilitate comparison with other algorithms.
MNIST handwritten data set is a set of digital data, each picture is 0 to 9, a single number, such as the following several:
MNIST database sources is a mixture of the two databases, one from the Census Bureau employees (SD-3), from a high-school students (SD-1); training sample has 60,000, 10,000 test samples. Training and testing samples, employee and student are written in each half. 60000 training samples a total of about 250 people to write. Crowd-sourced training and test samples of no intersection. MNIST database also retained handwritten correspondence between the numbers and identities.
TensorFlow dataset using MNIST
We can use Tensorflow provided input_data.py scripts to load data sets:
Author: secret_liang
link: https: //www.jianshu.com/p/29b14426a409
Source: Jane book
Jane book copyright reserved by the authors, are reproduced in any form, please contact the author to obtain authorization and indicate the source.