nano-neuron

Why we included this project

Most explanations of neural networks bury you in frameworks before you ever see a weight update. This one strips that away: a single neuron in about seven plain JavaScript functions, no libraries, no datasets, and it learns something real, converting Celsius to Fahrenheit, starting from a guess. The code shows prediction, cost calculation, forward and backward propagation, and the training loop end to end, so the math behind learning is right there in front of you instead of hidden inside an API call. For developers who want to feel how gradient descent works before moving to heavier tools, or teachers who need a self-contained demo a beginner can actually follow, it is a compact and honest place to start.