What you'll learn

Coding neural networks from scratch using only Python
What backpropagation is and how it helps machines learn
How to break down complicated math into simple, doable steps
The easiest way to understand gradients and why they matter
What's really happening when a machine makes predictions
How to train a smarter model by adjusting tiny details in code
This course strips neural networks to their fundamental core: math and raw Python.
You'll dive into the inner workings of backpropagation, gradient descent, and the math that powers modern neural networks. No pre-built frameworks, no black boxes. Just you, the math, and your code.
Step-by-step, you'll build neural networks by hand and implement them from scratch. From partial derivatives to weight updates, every concept is broken down and coded in Python (no libraries like PyTorch required!). If you're looking to truly understand how machine learning works—and prove it by building your own neural network—this course is your launchpad.
The course is broken down into three main sections
Introduction
Start by understanding the goals of the course and why backpropagation is central to modern machine learning. This section sets expectations and explains how mastering the math will give you a competitive edge.
Foundational Concepts and Simple Neural Network Implementation
Get hands-on with the theory. Learn how neural networks process data, calculate losses, and update weights using gradient descent. You'll manually compute everything—forward pass, gradients, and backpropagation—before coding a working network in Python.
Advanced Neural Network Implementation
Scale up your skills. This section walks you through implementing a deeper neural network with non-linear activation functions. You'll use advanced backpropagation techniques to train more complex models and understand how real-world neural networks are built from the ground up.
What Else Should I Know?
Homepage:

https://zerotomastery.io/courses/neural-network-from-scratch/