For Heads-of · Practitioner
Neural Network
A computational model loosely inspired by biological neurons, made of interconnected layers of artificial nodes that learn to recognise patterns in data.
- neural networks
- deep learning
- AI architecture
Structure
A neural network has an input layer (the data), hidden layers (which learn features) and an output layer (the prediction). The connections between nodes carry weights that are adjusted during training.
Deep learning
Deep neural networks, with many hidden layers, can learn complex layered representations of data. They are behind modern advances in image recognition, speech and natural language processing, including large language models.
Governance considerations
A trained network's reasoning is spread across millions or billions of weights, so it is difficult to interpret directly. That creates challenges for explainability and accountability, and for the people its decisions affect: the inadequate explainability for affected individuals risk.
Related Terms
Machine Learning
A branch of artificial intelligence in which systems learn patterns from data rather than being explicitly programmed for every scenario.
Model Interpretability
An intrinsic property of a model: the degree to which a person can understand the mechanism by which it turns inputs into outputs. A linear model or a shallow decision tree is interpretable by construction. A deep network is not, however good an explanation you can construct for it afterwards.
Transformer
A deep learning architecture that uses self-attention to process sequences of data, and the foundation of modern large language models.