https://github.com/joshsizer/mnist-digit-classifier

Keras, Pytorch, and Numpy only machine learning implementations for the classic challenge of classifying handwritten digits.
https://github.com/joshsizer/mnist-digit-classifier

Last synced: 3 months ago
JSON representation

Keras, Pytorch, and Numpy only machine learning implementations for the classic challenge of classifying handwritten digits.

• Host: GitHub
• URL: https://github.com/joshsizer/mnist-digit-classifier
• Owner: joshsizer
• License: mit
• Created: 2021-04-15T20:51:08.000Z (about 5 years ago)
• Default Branch: main
• Last Pushed: 2022-03-16T01:31:22.000Z (over 4 years ago)
• Last Synced: 2025-12-30T02:44:03.654Z (6 months ago)
• Language: Python
• Homepage:
• Size: 29.4 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# mnist_digit_classifier

Keras, Pytorch, and custom (from scratch) implementations.

The from scratch implementation is a 3 layer

network (784, 128, 10). It is currently at 97% accuracy.

Two things I could change to improve is swapping the loss

function from mean squared error to cross-entropy,

and the output layer activation from sigmoid to softmax.