MNIST dataset

As explained in Bayesian Optimization in Action

The MNIST dataset is a labeled collection of handwritten digit images commonly used for machine-learning training and evaluation.

The MNIST dataset contains labeled images of handwritten digits. Each image is a 28-by-28 matrix of integer pixel values from 0 to 255; zero represents white and 255 represents dark. Although digit recognition is classification, the numeric labels can also serve as continuous regression targets. section 13.1

A practical MNIST workflow represents and prepares the data as follows:

• Load the images with PyTorch and torchvision, convert them to tensors, and normalize pixel values using mean 0.1307 and standard deviation 0.3081. section 13.2.2 section 13.2.4
• Flatten each 28-by-28 image into 784 features while retaining its digit label as the target. section 13.2.2 section 13.2.4
• Train and test a Gaussian process by selecting 1,000 training images and 500 test images, then evaluate mean absolute prediction error. section 13.2.4

Figure 13.2 A data point from the MNIST dataset, which is an image with 28 rows and 28 columns of pixels, represented as a PyTorch tensor

Inspect a representative digit as both a 28-by-28 image and a PyTorch tensor of pixel values. section 13.1 section 13.2.2

How Bayesian Optimization in Action covers it

How deep it goes

Defines pixel structure, demonstrates tensor preparation, and works through a 1,000-image Gaussian-process experiment.

Examples

Shows a handwritten five, a 784-element tensor, and a train-test experiment using 1,000 training and 500 test images.

What you can do after

Apply it via PyTorch and torchvision to load, normalize, flatten, visualize, and split handwritten-digit data for GP prediction.

How it connects

Connects structured data, image classification, regression, Gaussian processes, PyTorch tensors, normalization, and RBF kernels.

liveBookmanning.com

Read Bayesian Optimization in Action on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Deep Learning with Python, Third EditionA standard dataset of handwritten-digit images used for machine-learning experiments and examples.
• 

  liveBookmanning.com
  Inside Deep Learning: Math, Algorithms, ModelsA classic handwritten-digit image dataset of 28×28 grayscale examples used here to illustrate image loading and tensor structure.
• 

  liveBookmanning.com
  Deep Learning with R, Third EditionMNIST is a dataset of grayscale 28-by-28 pixel images of handwritten digits labeled from 0 through 9.
• 

  liveBookmanning.com
  Deep Learning with JAXA standard handwritten-digit image dataset used for machine learning experiments and evaluation.

Compare how 5 books explain MNIST dataset →

Related concepts

• language modeluses · A language model is a computational model that represents patterns in linguistic tokens to estimate sequence probabilities and predict or generate language from context.
• benchmarkkind of · A benchmark is a defined task, dataset, workload, procedure, or reference measure used to assess or compare performance or capability under specified criteria or conditions.
• logistic regressionused by · Logistic regression is a statistical classification model that estimates categorical outcome probabilities by applying a logistic or softmax function to a linear combination of input features.
• image recognitionused by · Image recognition is the computational task of identifying or classifying objects, patterns, scenes, or other meaningful content in images.
• Fashion-MNISTalternative to · Fashion-MNIST is a labeled dataset of grayscale images representing ten categories of clothing and accessories for machine-learning training and evaluation.
• LeNetused by · LeNet is an early convolutional neural-network architecture designed primarily for handwritten-digit and character recognition.

This explainer was generated by Manning's knowledge graph from the text of Bayesian Optimization in Action. The section links point to the parts of the book it draws on.