Skip to content

dsgiitr/VisualML

Repository files navigation


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.

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