Lesson 9 of 12 · All lessons in this module
- Fintech and the four Vs of Big Data
- Sources of Big Data and its challenges
- Financial data science and what makes financial data hard
- AI and how machine learning learns
- Splitting data for machine learning: size, leakage and time order
- Overfitting, underfitting and model fit
- Supervised, unsupervised and deep learning
- Reinforcement learning and the four learning approaches compared
- Inside a neural network: layers, backpropagation, GANs and VAEs
- Data science: processing and visualisation
- Text analytics, NLP and data science tools
- Large language models and generative AI in finance
This lesson is part of the 2027 curriculum. You are following the 2026 curriculum, where it is not taught in this form. Switch if you are sitting the exam under the 2027 curriculum.
Lesson 9 of 12 · 12 min
Inside a neural network: layers, backpropagation, GANs and VAEs
A neural network passes data from an input layer through hidden layers to an output layer, learns by measuring its loss and pushing the error back through the network to adjust weights by gradient descent over many epochs, and the same building blocks power GANs and VAEs.
In short
- Input layer: one neuron per feature. Hidden layers: transform the data step by step (many hidden layers = a deep network). Output layer: a class or a continuous value.
- Layers are linked by weights, which adjust as the network learns.
- Training loop: forward pass → loss (prediction error) → backpropagation computes how each weight affects the loss (gradients) → gradient descent moves weights opposite to the gradient → repeat for many epochs.
- GANs: a generator and a discriminator network compete to produce realistic synthetic data, e.g., market scenarios for stress tests.
- VAEs: an encoder compresses data to a compact form and a decoder reconstructs it; used to reduce dimensionality and find hidden patterns or anomalies.
Unlock this lesson free for 7 days
Create a free account to get 7 days free: every lesson, video and flashcard, and the question bank. No card needed.