Exploring Building Topology Through Graph Machine Learning

Instructors: Wassim Jabi¹, Abdulrahman Alymani²

¹ Welsh School of Architecture, Cardiff University, U.K.

² Alfaisal University, UAE

Emails: JabiW@cardiff.ac.uk; abalymani@alfaisal.edu

Duration: 2 days

Mode of participation: Hybrid

Brief Description

This workshop integrates advanced spatial modelling and analysis and artificial intelligence, highlighting the importance of technological advancements in shaping the future of architecture and design. It introduces participants to novel workflows that link parametric 3D modelling with concepts of topology, graph theory, and graph machine learning. We will use TopologicPy, an advanced spatial modelling and analysis software library designed for Architecture, Engineering, and Construction, paired with DGL, a powerful machine learning library that provides tools for implementing and optimizing graph neural networks. Participants will learn how to use these tools to convert 3D models into graphs, analyze their properties, and perform classification and regression tasks.

The workshop begins with a brief introduction to the basics of graph theory, providing an overview of graph structures and terminologies such as nodes, edges, and graphs, as well as the properties that can be derived from these structures. Participants learn how to use TopologicPy to extract graphs from 3D models, and how to visualize and manipulate these graphs using platforms like grasshopper and Python scripting in Jupyter notebook.

The workshop will then cover more advanced topics such as centrality measures and shortest paths. Participants will learn how to use these algorithms to analyze and visualize graphs and derive relevant insights. They will also learn how to use unsupervised clustering techniques to group nodes with similar properties, which can be useful in a variety of applications.

We will then move on to machine learning with graphs, introducing the concept of graph neural networks and how they can be used to perform tasks such as graph and node classification and regression. Participants will learn how to use TopologicPy and DGL to create synthetic datasets based on generative and parametric workflows and build and optimize graph neural networks for specific tasks.

Throughout the workshop, participants will work on a variety of hands-on exercises and projects in small groups and will be guided by experienced mentors who are experts in the field of 3D modelling, graph theory, and machine learning. By the end of the workshop, participants will gain a deep understanding of the relationships between parametric 3D modelling, graph theory, and its application in machine learning. They will also learn how to use TopologicPy and DGL to represent, manipulate, and analyze graphs derived from 3D models.

Program

Day 1 – Full Day

(AM: Lectures, PM: Hands-on Workshop). Topics:

• Introduction to Topologicpy
• Introduction to Graphs
• Supervised Graph Machine Learning (Classification and Regression)
• Building a Synthetic Dataset

Day 2 – Half Day

(Hands-on Workshop, Documentation, and Discussion)

• Hyperparameter Optimization
• Training, Validation, Testing
• Testing on Unseen Data

Required Skills, Hardware, Software

Participants should have basic competency with Grasshopper or Blender and python programming. Examples, code snippets and close supervision will be provided.

This workshop will need only laptops with the required software installed. The needed software packages are:

• Windows OS
• Rhino 8 / Grasshopper
• Blender 3.X
• Python 3.10
• Jupyter Notebook, Jupyetlab, VSCode, and/or Google Colab
• Topologicpy
• Pytorch + DGL
• Plotly

All instructions will be shared ahead of time. Apart from Rhino 8 / Grasshopper and the Windows OS, all other software is free to download.