Machine learning and generative AI has become a game-changer for solving design problems. Besides well-known applications like drug discovery, machine learning can also greatly enhance AI-assisted design in architecture, civil and mechanical engineering.
Event source: FCL · Category: Public
In this workshop we will present how to create and deploy custom Conditional Variational Autoencoder (CVAE) models for design in architecture and engineering. We leverage capabilities of CVAE to perform forward and inverse design, and of the parametric modelling conventionally employed to design buildings, structures or parts. For forward design, the CVAE serves as a computationally efficient surrogate, streamlining the computational process through sensitivity analysis and uncertainty quantification. In inverse design, the user can specify the desired attributes of a parametric problem, and the CVAE proposes possible design solutions, supporting the designer in exploration of the design space.You will learn how to train and evaluate project-specific models, and deploy them to generate designs with the properties you request! For this, we will use our open-source toolkit for AI-eXtended Design (AIXD) which you can find here: https://gitlab.renkulab.io/ai-augmented-design/aixd.
Prerequisites:
- Attendees should bring their own laptop, and preferably pre-install the AIXD tools and Python environments required.
- The workshop is best suited for practitioners and scientists with at least intermediate experience in Python. It is aimed at architects and engineers with basic coding skills that want to leverage machine learning in their work, as well as computer/data scientists interested in applying their skills in the architecture or engineering domain.
* Recommended: basic understanding of machine learning, generative AI, experience with Python, jupyter notebook, pandas.