What is deep learning? Learning representations through layers
Deep learning is a branch of machine learning that uses neural networks with multiple layers to learn useful representations of data. It is used for tasks involving images, language, audio and other complex inputs. Depth refers to the layered computation, not human-like understanding.
SchoolWhool's current learning path covers Python, Data Tools, Maths, Statistics and classical Machine Learning. Open lessons' parts and key terms are free to read.
Why learning a representation matters
A model needs a numerical representation of its input. With a table, you may already have useful columns such as price and quantity. With a photograph, the starting point is pixel values. A layered network can learn intermediate representations rather than relying entirely on features chosen by hand.
Consider an exercise that sorts photographs of reusable containers by type. The interesting question is whether the model learns something about containers or merely the background used for each class. Keeping the task concrete makes it easier to design a test than starting with a network architecture.
How a deep learning model learns
Make a prediction
The input passes through layers with adjustable weights. The final output is interpreted according to the task, such as scores for image categories.
Measure the loss
A loss function compares the output with the training objective. A small training loss describes that objective on the training examples; it does not establish success on new inputs.
Compute gradients
Backpropagation applies the chain rule to calculate how the loss changes with the parameters. It supplies gradients rather than choosing the update on its own.
Update and evaluate
An optimiser uses those gradients to update weights. Check performance on separate data while developing the model, then reserve a final test for the chosen approach.
Where different architectures enter the picture
When is deep learning a sensible choice?
Treat it as a candidate to compare against a baseline. For the container exercise, define a simple reference method and test both on the same held-out images. Record training effort, prediction cost and the examples each approach misses. Avoid claiming that depth alone improves the result.
Before collecting more images, inspect your existing split. Photos of the same object taken seconds apart can make a random split misleading. Keep related images together, and try a separate lighting condition or background as a clearly labelled stress test.
A practical route into deep learning
Write and debug Python
Be comfortable with functions, loops, modules and reading an error message. Start with SchoolWhool's Python track if these are still unfamiliar.
Understand arrays and the maths
Practise array shapes and matrix operations, then connect derivatives and the chain rule to a loss function. The Data Tools and Maths tracks build these foundations.
Evaluate a classical model
Use the Statistics and Machine Learning tracks to practise splits, baselines and error analysis before adding a neural network.
Run one small experiment
Choose a modest dataset, make one change at a time and keep a log of what changed. Explain the evaluation before presenting the model's predictions.
Where this guide takes you
- Use the examples to understand the task and its evaluation. They are suggested learning exercises, not measured results or production-ready systems.
- The linked SchoolWhool courses teach the foundations. Specialist deep learning, computer vision, NLP, generative AI and reinforcement learning tracks are not published.
- Open lessons' parts and key terms are free to read. A free account opens quizzes and saved progress, plus practice and other material in the free sample modules.
Questions
What is the difference between machine learning and deep learning?
Deep learning is part of machine learning. It uses layered neural networks to learn representations; classical ML includes methods such as linear models, trees and nearest neighbours.
Do I need a GPU to learn deep learning?
Small networks and learning exercises can run on a CPU. Larger experiments may need an accelerator; choose the task before choosing hardware.
Does deep learning always need labelled data?
No. Deep networks can be trained with supervised, self-supervised and other objectives. Your evaluation still needs evidence appropriate to the intended task.
Is a neural network the same as a human brain?
No. The terminology is inspired by biology, but an artificial neural network is a mathematical model. Its predictions do not establish human-like understanding.
Does SchoolWhool offer a deep learning course?
A dedicated Deep Learning track is not published. The Python, Data Tools, Maths, Statistics and Machine Learning tracks teach foundations for studying it.
Sources
Technical references and further reading from the authors, educators and maintainers below. Read dates record when these references were checked; the examples and suggested exercises in this guide are SchoolWhool's own.
- Google Machine Learning Crash Course: neural networksRead 2026-10-14
- PyTorch: build the neural networkRead 2026-10-14
Related reading
Neural networks explainedWhat is computer vision?Deep learning study roadmapDeep learning tutorials: study orderMaths foundationsBuild the foundations for this subject
Follow SchoolWhool's learning path through Python, Data Tools, Maths, Statistics and Machine Learning. Start with the skills you need, read open lessons without an account, then sign in free for quizzes and saved progress. Free sample modules also open their practice and study material.
