Rocky playground

This repo is a fork of Agile's (archived) fork, which was a fork of Google's original project.

The app is currently deployed at playground.scienxlab.org.

About

Rocky Playground is an interactive visualization of neural networks, written in TypeScript using d3.js. It is a fork of the Google Neural Network Playground, and also incorporates code from David Cato's fork of the same project.

Some examples of what is different from the Tensorflow.org implementation:

• More activation functions including ELU, Leaky ReLU, and Swish.
• More loss functions, including hinge loss, log loss and Huber loss.
• More regularization (penalty) functions: elastic net and Huber.
• Added evaluation metrics.
• Real-world datasets related to geological tasks.
• Some new synthetic datasets: moons and linear (from matplotlib), and diagonal (to illustrate overfitting).
• The forward pass of the network is expressed as a Python function.
• You can change regularization on the fly, during training.
• You can upload your own datasets.
• Some bug-fixes and cosmetic changes.

Contributing

Want to help? We'd love to have your involvement! If you'd like to contribute, take a look at the contribution guidelines.

Development

You will need to install Node for your system.

Then, to run the visualization locally, run:

• npm i to install dependencies
• npm run build to compile the app and place it in the dist/ directory
• npm run serve to serve from the dist/ directory and open a page on your browser.

For a fast edit-refresh cycle when developing run npm run serve-watch. This will start an http server and automatically re-compile the TypeScript, HTML and CSS files whenever they change.