Oxford MSc AI for Business: Deep Neural Networks – Architectures & Backpropagation (2026)

UNIVERSITY OF OXFORD | MSC AI
FOR BUSINESS
MODULE: DEEP NEURAL NETWORKS (DNN)
TERM 1 LECTURE NOTES: ARCHITECTURES, BACKPROPAGATION & DEEP
LEARNING




1. FROM NEURONS TO NETWORKS
Deep Learning is a subset of Machine Learning inspired by the biological structure
of the human brain. At its core is the Artificial Neuron.
1.1 Anatomy of an Artificial Neuron
Each "node" in a network performs a simple mathematical operation:
1. Input (x): The data entering the neuron.
2. Weights (w): These represent the "strength" of the connection. Learning in
AI is simply the process of finding the right weights.
3. Bias (b): An extra parameter that allows the model to shift the activation
function.
4. Activation Function: A non-linear function (like ReLU or Sigmoid) that
decides whether the neuron should "fire."




2. MULTI-LAYER ARCHITECTURES
A network becomes "Deep" when it has more than one Hidden Layer.
• Input Layer: Receives the raw data (e.g., pixel values of an image).
• Hidden Layers: Where the "magic" happens. Each layer extracts
increasingly complex features.