Date of this Version
Chase, Andrew. 2019. “Machine Learning in Architecture: Connectionist Approach to Architectural Design.” Master's Thesis, University of Nebraska-Lincoln.
Previous applications to design processes intend to enhance a building’s schematic design using quantitative data. Therefore, most applications to the early design phases are passed by as simple overarching ideas informed by the designer and users’ knowledge. Although this is a preferred method of choice making, the knowledge used to inform conceptual and schematic design process can be limited. With the increase of computation in all major industries, a new increase in data to describe forms of infrastructure is required. These forms being objects to analyze their performance and potential, and active forms that can describe the disposition of urban space. Previous research into data driven design has worked its way into standardization and performance goals. However, the connection between active and object forms as a network of standards has yet to be introduced as a method of advising design. Meaning design has focused on creating a single object that seems to benefit the ecology. No attempt at connecting urban active data to object data has been made to benefit both forms equally, but only to preserve the status quo. The introduction of Machine learning algorithms has the capability to connect these complex forms and inform new designs. The application of machine learning algorithms advises the early phases of architectural design process by reviewing a manifold of data, privileging complex analogical connections, and simulating the designers informed symbolic choices.
Supervisor Professor Mark A. Hoistad, AIA