Last edited 26 Jun 2024

Generative adversarial networks

The 'CIOB Artificial Intelligence (AI) Playbook 2024', glossary of terms, published by the Chartered Institute of Building(CIOB) describes generative adversarial networks as a driver of recent AI developments.

'These are made up of two sub artificial neural networks: a generator network and a discriminator network. The generator network is fed training data and generates artificial data based on patterns in training data. The discriminator network compares the artificially generated data with the ‘real’ training data and feeds back to the generator network where it has detected differences. The generator then alters its parameters. Over time the generator network learns to generate more realistic data, until the discriminator network cannot tell what is artificial and what is ‘real’ training data and the AI model generates the desired outcomes. See also artificial neural networks and transformers.'

See generative AI, large language models and The long expanding list of AI tools for building planning, design, construction and management.

[edit] Related articles on Designing for Buildings

Designing Buildings Anywhere

Get the Firefox add-on to access 20,000 definitions direct from any website

Find out more Accept cookies and
don't show me this again