Skill
Neural networks
Layers, activations, backprop as computational graph. Intuition before frameworks.
Do these first
What you will be able to do
- Draw a 2-layer network
- Explain what a loss is optimizing
- Relate gradient descent to walking downhill
Curated resources
Start with the pick, then branch out
Scores are LearnPath editorial opinions (clarity, cost, freshness, project density) — not an objective ranking. Every YouTube item is embedded officially after a click-to-load facade.
1h 20mLearnPath pickLearnPath pickplaylistbeginner
Neural networks — Essence of neural networks
3Blue1Brown1h 20m
Watch the whole playlist. Intuition first, frameworks second.
Score 99
Watch
6hProject-basedAlternativeyoutube coursebeginner
Machine learning with Python — full course
freeCodeCamp / freeCodeCamp.org6h
sklearn-oriented. Pair with official User Guide.
Score 87
Watch
1h 26mFastest pathQuick pathyoutube videobeginner
Deep Learning Crash Course for Beginners
freeCodeCamp.org1h 26m
Eighty-six minutes that cover the main architecture families — dense, convolutional, recurrent — with code, so you see what each is for before committing to a framework course.
Score 73
Watch
1hOfficial docsOfficial docsyoutube courseadvanced
Stanford CS230 | Autumn 2025 | Lecture 1: Introduction to Deep Learning
Stanford Online1h
Stanford CS230's current (autumn 2025) opening lecture: what deep learning is, how the course builds, and the practical advice on data, splits and metrics that the lectures emphasise. The academic anchor for a self-taught deep-learning path.
Score 78
Watch10 videosLearnPath pick
More curatedplaylistintermediate
Neural Networks: Zero to Hero
Andrej Karpathy19h 21m
Ten videos building neural networks from scratch — backpropagation with micrograd, language models with makemore, then the GPT build. Around nineteen hours total and the most rigorous free path from intuition to implementation.
Score 84
Open