this post is about some multi layer perceptron online calculators with javascript/typescript for educational purposes
One of the most famous: Tensorflow Playground
1/NIST calculator (release 2020) has a lot a features from Tensorflow Playground (release 2018).
i luv so much this online service:
The GitHub deployment and repositories of Neural Network Calculator can be found at:
GitHub nn-calculator code: https://github.com/usnistgov/nn-calculator
GitHub deployment code: https://github.com/usnistgov/nn-calculator/tree/nist-pages
GitHub nn-calculator deployment: https://pages.nist.gov/nn-calculator/
2/Tensorflow Playground group wrote a tiny neural network library that meets the demands of educational visualization.
Source code: https://github.com/tensorflow/playground
Deep playground is an interactive visualization of neural networks, written in TypeScript using d3.js.
3/TensorFlow was developed by the Google Brain team.
The initial version was released in 2015. Google released the updated version of TensorFlow, named TensorFlow 2.0, in September 2019. In 2023 Feb the release is 2.12. https://github.com/tensorflow/tensorflow
4/Tensorflow Playground is also based on some ideas of precursors:
the perfect Chris Olah’s articles about neural networks (2014):
And the work of karpathy when he was a PhD student at Standford (release 2014)…
https://cs.stanford.edu/people/karpathy/convnetjs/
a very simple perceptron calculator
an old online interactive book in english and in french (release 2014)
https://mlweb.loria.fr/book/en/perceptron.html
another simple perceptron calculator
The plan:
We're going to train a single neuron (perceptron) to linearly separate some 2D points.
All done in pure JavaScript, from scratch with no libraries
release 2021
A metric for how well the weights worked
Minimise this metric until it's good enough...