- New
- Intermediate
- Text, code and quiz
Neural Networks from Scratch
Build a neuron, the ReLU and sigmoid activations, a loss function and gradient descent in plain Python, and watch a network learn.
- Level
- Intermediate
- Study time
- 1h 3m
- Chapters
- 2
- Lessons
- 5
- Quizzes
- 1
Created by ReLU Learning Team
Courses by the ReLU Learning team · Last updated October 2026
About this course
Deep learning frameworks hide the mechanics. Here you write them yourself: a neuron as a weighted sum, activations, a loss, and the gradient descent loop that every neural network is trained with.
No libraries and no hand-waving. Every step is code you run in the lesson, small enough to understand line by line.
What you will be able to do
- Compute the output of a neuron by hand and in code
- Explain what ReLU and sigmoid do and why networks need them
- Minimise a loss with gradient descent
- Train a neuron to learn a logical function
What you need first
- Comfortable Python
- Machine Learning Foundations, or equivalent
Course content
2 chapters · 5 lessons · 1 quiz
1. One neuron
2 lessons · 22m
- TextA neuron is a weighted sumFree preview10m
- TextActivation functions, ReLU included12m
2. Learning
3 lessons · 41m
- TextLoss and gradient descent15m
- CodeTraining a neuron on AND20m
- QuizCheck your understanding6m
About the instructor
ReLU Learning TeamCourses by the ReLU Learning team
- 4 courses
- 1 learner
Hands-on courses in Python, machine learning and AI, written so that every idea is something you can run.