https://github.com/dkealvaro/mnist

Achieved 75% accuracy in the MNIST dataset using a simple cosine similarity approach under 50 lines of code, with no training
https://github.com/dkealvaro/mnist

cosine-similarity machine-learning mnist numpy

Last synced: 3 months ago
JSON representation

Achieved 75% accuracy in the MNIST dataset using a simple cosine similarity approach under 50 lines of code, with no training

• Host: GitHub
• URL: https://github.com/dkealvaro/mnist
• Owner: DKeAlvaro
• Created: 2025-03-05T09:13:07.000Z (7 months ago)
• Default Branch: main
• Last Pushed: 2025-03-05T09:50:03.000Z (7 months ago)
• Last Synced: 2025-03-05T10:39:51.111Z (7 months ago)
• Topics: cosine-similarity, machine-learning, mnist, numpy
• Language: Python
• Homepage:
• Size: 0 Bytes
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          


    



# MNIST Digit Recognition

Achieved 75% accuracy using a simple cosine similarity approach under 50 lines of code, with no training.

Each instance from the test set is compared with `num_neighbors` instances of each digit from the training set, and the digit with the highest sum of similarities is predicted. As we can see in the plot below, the accuracy increases with the number of neighbors.