| Speaker(s) | Rossella Arcucci Imperial College London |
| Date | 27 April 2023 – 09:00 to 10:00 |
| Venue | INI Seminar Room 1 |
| Session Title | Data Learning: integrating Data Assimilation and Machine Learning to deal with limitations in models and data |
| Chair | Andrew Duncan |
| Event | [DDEW03] Computational Challenges and Emerging Tools |
| Abstract | Computational modelling of real world systems is intrinsically limited by the many types of imperfections encountered when training systems on real world data. Machine Learning (ML) for physical systems is the approximation of the system based on this data, and involves the prediction of physical attributes and dynamics from this data. Data Assimilation (DA) is the correction of the approximation of a physical system by the integration of observations with a dynamic model. This allows for more accurate temporal models to be developed, by correcting the temporal slip and non-physical predictions made by these models. We outline a range of methods which are currently being implemented involving ML and DA, and discuss how these mitigate the various problems associated with training models with data. From this, we demonstrate that a variety of methods are currently being used in the field of Data Learning, and illustrate improved results in a spread of different real world applications. We offer a guide to how these methods may be implemented to deal with certain types and limitations in data. |
Data Learning: integrating Data Assimilation and Machine Learning to deal with limitations in models and data
Supported By
