Help us understand what motivates
engineers to retrain ML models
We are researchers from
Carnegie Mellon University
and Instituto Superior Técnico
looking for industry practitioners that:
- work for companies that leverage ML models in their products/services/pipelines
- periodically update or plan to start updating their company's ML models in the near future
- have some influence in the ML pipeline, even if their role is not traditionally associated with ML
We are interested in understanding:
- what triggers a ML model update, how often ML models fail and what causes those failures
- how often ML models are updated and how expensive that process is
- whether perceptions of system and ML model goals differ depending on a practitioner's role
- what practitioners feel is missing regarding tools to enable periodic ML model updates
Meet our team
Maria da Loura Casimiro
PhD Candidate at the Software and Societal Systems Department (S3D), Carnegie Mellon University and at Instituto Superior Técnico, University of Lisbon.
Paolo Romano
Professor at Instituto Superior Técnico, University of Lisbon and senior research at INESC-ID.
David Garlan
Professor at the Software and Societal Systems Department (S3D), Carnegie Mellon University.
Luís Rodrigues
Professor at Instituto Superior Técnico, University of Lisbon and senior research at INESC-ID.
Christian Kästner
Professor at the Software and Societal Systems Department (S3D), Carnegie Mellon University.
Previous work on ML model retrain
Towards a Framework for Adapting Machine Learning Components
Self-adaptive Machine Learning Systems: Research Challenges and Opportunities
A Probabilistic Model Checking Approach to Self-Adapting Machine Learning Systems
