MLX + MNIST: Build a Neural Network from Scratch on Apple Silicon

In this tutorial, we train a Multi-Layer Perceptron (MLP) using Apple’s MLX framework and the MNIST dataset. Whether you're new to deep learning or want to explore MLX on Apple Silicon, this hands-on guide walks you through the full process—from model architecture and backpropagation to performance evaluation.

🔧 What you’ll learn:
What a multi-layer perceptron is
How to implement and train it using MLX
How to prepare the MNIST dataset
How backpropagation works
How to evaluate a multi-label classification model

💡 Tools used:
Apple MLX (optimized for M1/M2 chips)
MNIST dataset (handwritten digit classification)

Python
📌 No Mac? You can still follow along with the PyTorch version of the code—linked in the description!

👨‍🏫 This video is part of a mini-series on AI, physics, and finance. Subscribe for more practical tutorials that bridge theory and implementation!

📁 GitHub Code Repository:
👉 github.com/espitia01/EntroPhys/tree/main/neural-networks/mlps-1

Видео MLX + MNIST: Build a Neural Network from Scratch on Apple Silicon канала EntroPhys