Architecture Engineers Award

The increasing complexity of architectural layout design has highlighted the need for computational models capable of understanding relational spatial structures. Despite graphs being a natural fit for representing spatial dependencies, research in graph-based interpretation and generation of architectural layouts remains limited. SE-VGAE addresses this gap by introducing an unsupervised disentangled graph representation learning framework tailored for architectural layout design. The method focuses on capturing complex spatial relations through attribute-rich multigraphs, enabling deeper understanding and generative capability in design automation. Graph-Based Representation Learning in Architectural Layouts Architectural layouts inherently contain spatial dependencies, adjacency relations, and functional hierarchies that can be efficiently modeled using attributed adjacency multigraphs. This topic explores how SE-VGAE leverages graph-based learning to encode layout features beyond...