https://github.com/kevinrobinson/tiny-trainer

experiment with using the universal sentence encoder and KNN to make a classifier from small training sets
https://github.com/kevinrobinson/tiny-trainer

Last synced: 4 months ago
JSON representation

experiment with using the universal sentence encoder and KNN to make a classifier from small training sets

• Host: GitHub
• URL: https://github.com/kevinrobinson/tiny-trainer
• Owner: kevinrobinson
• Created: 2019-03-22T01:41:26.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2019-04-24T14:11:00.000Z (over 5 years ago)
• Last Synced: 2024-06-22T15:55:18.406Z (6 months ago)
• Language: JavaScript
• Homepage: https://tiny-trainer.netlify.com/
• Size: 1.81 MB
• Stars: 6
• Watchers: 4
• Forks: 1
• Open Issues: 2
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# tiny-trainer

experiment with using the universal sentence encoder and KNN to make a classifier from small training sets

### Demo

https://tiny-trainer.netlify.com/

![dataset](docs/dataset.png)

![results](docs/results.png)

![t](docs/t.png)

![q](docs/q.png)

![ingredients](docs/ingredients.png)

### Using your models

You can click "Copy!" to get script tags you paste in an HTML page to make predictions in your own app.

![repl](docs/repl.png)

This is an example repl.it you can fork, see [https://repl.it/@kevinrobinson/WeeFancyTriangles](https://repl.it/@kevinrobinson/WeeFancyTriangles)

![own-app](docs/own-app.png)

### References

- https://github.com/tensorflow/tfjs-models/blob/master/universal-sentence-encoder/demo/index.js

- https://github.com/tensorflow/tfjs-models/tree/master/knn-classifier

- https://machinelearningforkids.co.uk

### Development

This is static website, but it uses `node.js` for the build process.

In the project directory, you can run:

```

$ yarn install # install dependencies

$ yarn start # run build in development mode

```

Then open [http://localhost:3000](http://localhost:3000) to view it in the browser.

This project was bootstrapped with [Create React App](https://github.com/facebook/create-react-app), so that's a good place to learn more if you are trying to set it up yourself.