MNIST Dense Neural Network 3D Visualizer

Minimal 3D view of a dense network learning MNIST in real time. Green/red edges show positive/negative weights, left panel shows structure and loss, right panel shows your drawing and predictions.

Why it matters: it turns abstract tensors into spatial intuition, helping learners connect architecture, weight updates, and loss trends at a glance.

The project uses JAX under the hood for all the ML math, and React + Three.js for the 3D rendering and layout.

Install & Run

Prerequisites

• Bun (for frontend)
• Python 3.10+ (for backend)

Frontend (Vite + React + three.js)

bun install
bun dev

App runs at http://localhost:5173.

Backend (FastAPI + Uvicorn)

cd backend
python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -e .
uvicorn app.main:app --reload --host 127.0.0.1 --port 8000

API runs at http://127.0.0.1:8000.

That’s it, keep both processes running while developing.

x/github/everything else: @xandykati98

Contact-me: [email protected]