What is MNIST (what does plist mean)

Hello everyone, meet again, I'm your friend Quanstack Jun.

Initial confusion

I think many rookies, like me, started to learn machine learning with zero basic knowledge.It is estimated that many people start to learn ML, and they are confused and have no idea what they are talking about.Because the learning mode is completely different from learning other languages, we know that when learning other languages, the first program is to print "Hello World".

In fact, machine learning uses a certain language to solve problems under a certain framework.Therefore, for zero-based rookies, we need to learn a certain language first. We can recommend Python because of its powerful functions and relatively simple syntax, and C++ can also be used.As for the framework, my personal recommendation is TensorFlow2, because Google's thighs are thick.

What is MNIST

It is recommended to start working under the framework of TF2 after understanding Python.

An introduction to machine learning is MNIST.The MNIST data set comes from the National Institute of Standards and Technology in the United States and is a reduced version of NIST (National Institute of Standards and Technology). The training set consists of handwritten numbers from 250 different people, 50% of which are high school students.50% were from the Census Bureau staff, and the test set was the same proportion of handwritten digit data.

Get MNIST

MNIST dataset is available at http://yann.lecun.com/exdb/mnist/Get, the picture is stored in the form of bytes, it contains four parts:

• Training set images: train-images-idx3-ubyte.gz (9.9 MB, 47 MB ​​uncompressed, contains 60,000 samples)
• Training set labels: train-labels-idx1-ubyte.gz (29 KB, 60 KB unpacked, contains 60,000 labels)
• Test set images: t10k-images-idx3-ubyte.gz (1.6 MB, 7.8 MB unzipped, contains 10,000 samples)
• Test set labels: t10k-labels-idx1-ubyte.gz (5KB, 10 KB unpacked, contains 10,000 labels)

In this dataset, there are 60,000 training samples, of which 55,000 are used for training and the other 5,000 are used for validation.Test samples: a total of 10,000 samples, with the same proportion of verification data.

Pixel values ​​in the dataset:

a) Use python to read the binary file method to read the mnist data set, the pixel value of the read image is between 0-255; the label is the value of 0-9.

b) Using TensorFlow's encapsulated function to read mnist, the pixel value of the read image is between 0-1; the label is a row vector with a size of 1*10 composed of 0-1 values.

Publisher: Full-stack programmer, please indicate the source: https://javaforall.cn/126875.htmlOriginal link: https://javaforall.cn