Parallel sessions (workshop) - Machine learning with sparse data

When production volumes are low or events are rare, there’s still too much data for a human to sift—and not enough for “brute‑force” machine learning. Industry often faces the “small but important” data regime: early-stage manufacturing and yield ramp-up, predictive maintenance for high-reliability assets with few failures, quality inspection with scarce defects, design‑of-experiments in R&D, and other domains where labels or events are limited (e.g. materials discovery, remote sensing, clinical and safety‑critical settings). This workshop explores how to extract signal from limited, noisy or imbalanced datasets to deliver reliable value in other data-sparse contexts.

The workshop consists of a few short thought-provocations from our experts, followed by facilitated breakout discussions with participants.

Moderator: Matthew Burdge

Contributors & title of talk(More details to be announced):

• Dr. Alex Serb

• Dr. Dan Hodgson

• Dr. Rosa Filgueira

With the workshop, we aim to:

• showcase University of Edinburgh's expertise in data, digital and AI

• surface cross-sector challenges where sparse data ML can create measurable benefit

• co-create opportunities for collaboration around new solutions or questions

We invite attendees to share their experiences, insights, and ideas to the workshops, bringing something to add or to ask to contribute to the breakout discussions.