This course introduces students to the fundamentals of Machine Learning (ML). By working in groups, the class develop an “architectural” case study, where they generate their own dataset, and train a model for a deployment case of their choice. This gives students hands on experience with ML, focussing on process and critically evaluating their workflows.
In a moment where architectural practice is increasingly saturated with AI-driven marketing, buzzwords, and opaque “black box” claims, it is becoming harder to distinguish genuine mathematical principles from technological hype. This course intentionally steps away from the noise to focus on fundamentals. As architecture enters the AI race, future practitioners need a clear, hands-on understanding of how these technologies actually work, not just how they are advertised. This seminar offers a scaffolded, practice-based introduction to the core mechanics of machine learning. Through best-practice examples, rigorous dataset curation, and simple yet powerful modelling workflows, students will learn how targeted ML predictions can be meaningfully applied to architectural questions. Emphasis is placed on designing custom datasets and understanding how their structure interfaces directly with model behavior, laying the foundation for reliable, interpretable, and domain-specific AI applications informed by architectural domain knowledge.
Here we don’t look at ML as “intelligence” but rather as an alternative data-driven modelling paradigm for problems that might be too complex for mechanistic computation. Differently from analytical or procedural models where processes are computed using iterative, time-based, heavy calculations and pre-assigned property parameters, machine learning models learn by example through searching for an approximated relationship between the inputs and the targets. Once trained, they can be easily deployed within a design workflow to offer targeted predictions. While these models promise advantages and productivity, they are notoriously data-hungry. Many open source datasets are available online for training of state-of-the-art models, however these datasets are not relevant for architectural applications.
Dataset design is the enabler to using Machine Learning algorithms in different fields. While an existing ecology of algorithms provides architectures suitable to specific types of problems be classification or regression, the models must be trained on datasets pertaining to the problem they are expected to solve. These custom datasets can be composed through web-scraping, or be generated, either computationally through heuristic algorithms, or through sensing and digitizing physical samples. Dataset design requires several criteria to be followed: Increased sample diversity, bias avoidance and ambiguity avoidance. These criteria ensure that the problem is well represented, with equal distribution and without confusion – and is the key to a successful ML training campaign.
In this seminar, you will be introduced to conceptual perspectives around dataset design, feature encoding, for targeted ML applications. You will develop your own ML pipeline for a use case of your choice taking departure from an architectural application of your interest. Working in groups, you will be tasked to curate a dataset and train an adequate model for your predictive task. Through the study of the parameter space and feature distribution and representation, you will propose a dataset encoding method, and evaluate it through the training of shallow models as well as artificial neural networks. A key milestone will be designing a computational pipeline bringing the prediction back into the design workflow. This is an occasion to step back from “mega-models” such as LLMS and Diffusion Models, and build a foundational understanding of the basics of ML.
At course completion the student will:
Tile-based machine-learning workflow for indoor plant suitability prediction Introduction Plant placement in interior spaces is usually treated as an intuitive or aesthetic decision. However, different areas inside the same room can receive very different levels of sun, radiation, useful daylight and humidity. This project proposes a machine-learning workflow to classify interior space into plant suitability … Read more
ML-Driven Plant Placement for Adaptive Architecture BioSpatial Intelligence explores how machine learning can support planting decisions in architectural spaces. The project starts from a simple design question: when we design a building, how can we decide which plants belong to which environmental conditions? Instead of relying only on intuition, we developed a workflow that reads … Read more
Objective Traditional flood risk maps take months to produce, are updated only every few years, and are too coarse, they might say a whole district is at risk without telling you which specific street or field will actually be underwater. “We are going to predict whether any given location in Thessaly, Greece is Flood-Prone or … Read more
Can we predict what kind of use does a city grid hosts — Commercial vs. Residential — from its built form, morphology, and proximity to other urban features? “Can we predict the land use of a space based on existing environmental information from official and unofficial sources?” What’s the sweet spot for a Machine Learning … Read more
