Neural Nets & Backprop in the Browser — Vanilla JS, No OOP, No Libraries!

Join me for a beginner-friendly walkthrough where we build and train neural networks — taking a functional approach, entirely in the browser using vanilla JavaScript. We’ll cover gradient descent, hidden layers, and backpropagation, and by the end, we’ll build an app that can classify mushrooms and handwritten digits! Neurotical: https://neurotical.vercel.app Check out the source code and follow along commit-by-commit: https://github.com/jmazin/neurotical Resources & Attributions: "Back Propagation Example" by Glosser.ca – https://commons.wikimedia.org/wiki/File:Back_Propagation_Example.svg (CC BY-SA 3.0) "1D Convolutional Neural Network feed forward example" – https://commons.wikimedia.org/wiki/File:1D_Convolutional_Neural_Network_feed_forward_example.png (CC BY-SA 4.0) Universal Approximation Theorem article – https://www.geeksforgeeks.org/universal-approximation-theorem-for-neural-networks/ Graphs created using Desmos – http://desmos.com Inspiration and excerpts from Sebastian Lague – https://www.youtube.com/watch?v=hfMk-kjRv4c