Skip to content

This repository contains a hyperparameter tuning experiment for a Convolutional Neural Network (CNN) trained on the MNIST dataset.

Notifications You must be signed in to change notification settings

brownbeardeveloper/hyperparameter_tuning_mnist_cnn

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 

Repository files navigation

hyperparameter-tuning-mnist-cnn

This project is about tuning hyperparameters for a Convolutional Neural Network (CNN) trained on the MNIST dataset.

I started with a grid search, testing 128 different parameter combinations. Each one was trained for just 1 epoch because of limited compute. Then I picked the top 5 setups based on validation accuracy.

These top 5 models are now trained longer, but not for a fixed number of epochs. Training stops if the model doesn't improve for 5 epochs in a row.

This is my first machine learning project, and it’s been interesting to see how much tuning can affect performance.

About

This repository contains a hyperparameter tuning experiment for a Convolutional Neural Network (CNN) trained on the MNIST dataset.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published