Intro to Deep Learning
About this free Intro to Deep Learning course
Intro to Deep Learning is a free deep learning course that explains the math behind neural networks, the models inspired by biological neurons that power Transformers, CNNs, and modern AI. Instead of treating AI as a black box, it shows how a network is really a mathematical function and how it learns by adjusting its weights to reduce error.
The course builds from real and artificial neurons to linear regression as a two-layer network, mean squared error loss, partial derivatives, and gradients, with exercises along the way. It then explains gradient descent, the iterative algorithm used when no exact solution is available. It finishes with auto-regression for time-series analysis, the idea behind forecasting and modern language models.
Who this course is for
This course is best for high school juniors and seniors and early college students who are comfortable with Algebra 2. Some exposure to calculus, especially derivatives, will help with the gradient sections. Intro to Machine Learning is a good warm-up.
Course outline
- What is a neural network? Domains, model types such as Transformers and CNNs, and algorithms and techniques.
- Neurons, biological and artificial. How real neurons inspired artificial neural networks, and why deep networks are called deep.
- Linear regression as a neural network. Viewing linear regression as a two-layer network and measuring error with mean squared error (MSE).
- Partial derivatives and gradients. The calculus that tells a model which way to adjust, with worked exercises.
- Training a 2-layer linear network. Solving for the weights, adding a bias term, and a one-dimensional example.
- Gradient descent. The iterative algorithm that trains nearly every modern neural network.
- Auto-regression. Single and multivariate auto-regression and its role in deep learning.
What you'll learn
- Describe a neural network as a mathematical function
- Compute mean squared error loss for a simple model
- Calculate partial derivatives and gradients
- Explain and apply gradient descent to minimize a loss function
- Describe how auto-regression is used to model sequences
How long it takes
Plan on about 8 hours. The math-heavy sections on gradients and gradient descent reward working the exercises by hand.
Frequently asked questions about Intro to Deep Learning
Is Intro to Deep Learning really free?
Yes. Intro to Deep Learning is completely free, with no fees, subscriptions, or account required. STEM for Others is a 501(c)(3) nonprofit, and every course in our curriculum library is free for students, families, and teachers.
How much math do I need for deep learning?
You'll need solid algebra and comfort with functions. The course introduces partial derivatives and gradients as it goes, but students who have seen derivatives in Intro to Calculus will find it smoother.
How do I start the course?
Click Begin Course at the top of this page to open the full slide deck in a new tab. You can also flip through the slides embedded above. Work through one unit at a time and try the practice questions before moving on.
Do I get a certificate?
The self-paced Intro to Deep Learning materials don't include an official certificate. If you're taking the course through a STEM for Others chapter program, ask your chapter leader about recognizing your completion.
Can teachers and clubs use this curriculum?
Yes. Teachers, homeschool families, libraries, and after-school clubs are welcome to use the Intro to Deep Learning materials with their students at no cost. STEM for Others volunteers also teach our curriculum in partner schools and libraries. You can find a chapter near you or start one at your school.
Related free courses
