Machine learning can be a powerful tool to aid in the creation of adaptive architecture models and developing interesting possibilities in design that consider principles of a circular economy.

Switching architecture to become a circular economy poses new challenges in how we organize construction, utilization and recycling of our living spaces. Architects will not only design the building as an end product but also its entire timeline – how it’s conceived, how it’s used and what will happen with the materials after disassembly. In this context, machine learning can aid the architects in a range of ways among others:

• Predicting the usage patterns (wear and tear, maximizing use phase)
• Responding to the occupant’s needs (efficiency improvements)
• Intelligent scrap material evaluation (addressing recycling)

During this workshop, the participants will get a chance to learn about machine learning in the design context, with the applications ranging from the clustering of elements to space planning applications for generating interior architecture responsive to user needs. For that purpose a parametric model will be developed, which will serve as a data source. Using unsupervised and supervised learning techniques, the participants will then perform a range of tasks addressing the model solution space. 

The first task will be to use clustering methods to discover families of possible designs. Dimensionality reduction using autoencoders will be used to select a subset of preferred solutions. Supervised learning will predict the indirect properties of the design based on user input. A survey will be designed to gather information about the user preferences regarding the design solution space, in an effort to prolong the usage phase and serve the users better (usage intensification). 

No prior knowledge of AI or machine learning is required. A basic understanding of Grasshopper is required.

• Supervised learning with neural networks
• Unsupervised learning with k-means clustering
• Data pre- and post-processing
• Data collection

• Rhinoceros 6 (free versions available)
• Grasshopper Plugins installed: Owl + Lunchbox + Shortest walk
• A second monitor is highly recommended

Basic understanding of Grasshopper is recommended (you can register to our Computational Design for Beginners workshop).

Each participant needs to work from their personal computer.

The principal software is Rhinoceros 6.0. The 90-day trial version can be downloaded from the website www.rhino3d.com/eval.html . Grasshopper 3D will be the main computational tool.

Windows operating system is highly recommended since many of the Grasshopper plugins still do not run on Apple IOS operating systems.