Skip to content

Neural Network teaching tool in the browser. Lets you visualize Neural Networks and individual training epochs.

Notifications You must be signed in to change notification settings

srikarboga/HackRU24

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

68 Commits
 
 
 
 
 
 
 
 

Repository files navigation

HackRU24 App - RESTLESS LEARNING

🎉Rutgers Hackathon 2024 Winner in the Education Track🎉

This app was built in 24hrs for the Rutgers Hackathon event. It is intended to help students visualize and play with neural networks and build intuition for how they can be trained.

Build instructions:

Make sure python and npm are installed before proceeding.

Clone this repository

Move to the frontend directory
cd ./frontend
Install npm dependecies
npm install
Move to backend directory
cd ../backend
Install python dependencies
pip install -r requirements.txt
Run backend server
uvicorn backend:app --reload

Then in a new terminal open the frontend directory

Run frontend development server
npm start
Finally, open http://localhost:3000 in a browser of your choice to interact with the website.

App Demo

The sliders can be used to adjust the layer size and number of layers of the neural network. The input and output layers cannot be changed. The model is then created on the backend.

Screen Shot 2024-03-24 at 11 50 36 AM

After each press of the button at the center the model, the model calls the backend server to run one training iteration on the MNIST dataset. The server returns the results the training of the iteration in terms of the accuracy of the model on the validation dataset, and returns this to the frontend to be displayed for the user.
image

There are also some helpful tips to build intuition for how neural networks learn that show up on the bottom left side of the screen after a certain number of epochs.

The goal of this is to let students build intuition for how neural networks work.

About

Neural Network teaching tool in the browser. Lets you visualize Neural Networks and individual training epochs.

Resources

Stars

Watchers

Forks

Contributors 3

  •  
  •  
  •