Build your first neural network from the ground up — understand neurons, layers and activations, see how forward pass and backpropagation actually work, and train a real model in Keras/TensorFlow. You will finish able to design, train and validate a simple network on both tabular and image data, and judge when deep learning is genuinely worth it.
Watch the free preview
From a single neuron to a network — free to watch, no account needed.
What you'll learn
- Build your first neural network and train it end to end in Keras/TensorFlow
- Explain neurons, layers and activation functions and choose them correctly
- Describe the forward pass, loss and backpropagation that let a network learn
- Train and validate a model, reading the loss curves to spot under- and over-fitting
- Decide when deep learning is worth it versus a simpler classical model
Syllabus
Neurons, Layers and Activations
From a single neuron to a networkFree preview
Activation functions and network shape
How Networks Learn
Forward pass and loss functions
Backpropagation and optimisers
Building and Training with Keras
Your first network in Keras
Training, validation and reading the curves
Deep Learning in Practice
An image example with Keras
When deep learning is (and isn't) worth it
Lab — 10 Exercises & Solutions
Exercises 1–5
Exercises 6–10