Visual Machine Learning

Visual Machine Learning contains a set of Machine Learning and Deep Learning interactive visualisation demos for developing intuition. These demos are developed using TensorFlow.js and can be executed directly in your browser. This project is an extension of ML examples from tfjs-examples. We implement new demos, as well as, add additional features into the ones that already existed in TFJS.

Some examples may require web-gl enabled browsers and viewers may experience latency during executing the demos based on the device.

Overview of Demos

Example name	Demo link	Input data type	Task type	Model type	Training	Inference
ANN	🔗	Iris Dataset	View NN architecture, View Confusion Matrix	Multilayer perceptron	Browser	Browser
Autoencoder	🔗	MNIST dataset	Visualising Latent Space	Autoencoder	Browser	Browser
Logistic Regression	🔗	Various 2D data	Visualising Decision Boundary	Logistic Regression	Browser	Browser
MNIST-CNN	🔗	MNIST	Visualising Activations	CNN	Browser	Browser
PCA	🔗	Various	Visualising Principal Components & projected dimensions	PCA	Browser	Browser
SVM	🔗	2D Dataset	Visualising Support Vectors and Kernels	SMO	Browser	Browser
Neural Style Transfer	🔗	Image Data	Visualising Style Transfer using MobileNet	Style Transfer	Browser	Browser
Vanishing Gradients	🔗	Iris Dataset	Developing Intuition how Relu Fixes Vanishing Gradients	Neural Networks	Browser	Browser

Dependencies

All the examples require the following dependencies to be installed.

Node.js version 8.9 or higher
NPM CLI OR Yarn

How to build?

cd into the directory

If you are using yarn:

cd MNIST-CNN
yarn
yarn watch

If you are using npm:

cd MNIST-CNN
npm install
npm run watch

Details

The convention is that each example contains two scripts:

yarn watch or npm run watch: This starts and generates a local development HTML server tracking filesystem for changes, supporting hot-reloading.
yarn build or npm run build: generates a dist/ folder which contains the build artifacts and can be used for deployment.

Contributing

If you want to contribute a demo, please reach out to us on Github issues before sending us a pull request as we are trying to keep this set of examples small and highly curated.

Acknowledgements

tfjs-examples