The Discovery AI module is organised around three connected discovery problems:

• Signature Discovery (Objects): how to discover what to measure—interpretable signatures, biomarkers, phenotypes, and latent states—using classical methods (e.g., PCA/NMF, clustering, feature selection) and modern representation learning (e.g., embeddings, VAEs).
• Causal Discovery (Relations): how to discover how variables relate—recovering plausible causal structure from observational, interventional, and temporal data under explicit assumptions, and understanding what is (and is not) identifiable (e.g., CPDAG/PAG outputs).
• Hypothesis Discovery (Systems): how to discover why the system works—turning signatures and causal graphs into testable mechanistic hypotheses, specifying discriminating predictions, and proposing minimal discriminating experiments within an auditable workflow.

Audience: NILAB PhD students and PhD researchers across the universities.

Duration: 30 hours

Format: concept sessions + practical capstones (code provided; heavy coding not required)

Room: Allstate Software Studio on the 2nd Floor of the Computer Science Building, QUB.

Time: 10am-4pm