Enriching Machine learning-based Pipelines with the Physics-inspired Features: Case Study of Thermal Processes and Systems

The present seminar is focused on the role of physics-inspired feature generation in improving the performance of machine learning-based pipelines and enhancing their physical interpretability. Simulating the thermal processes taking place in indoor environments and estimating the buildings' characteristics from smart meter data are specifically presented as case studies. The impact of selecting the most promising set of features on reducing the pipelines' dimensionality and facilitating the physical interpretation of the models is also discussed.
Speaker(s): Behzad,
Inndalsveien 28, Bergen, Vestfold, Norway, 5063, Virtual: https://events.vtools.ieee.org/m/370098