Knowledge for contextualization of machine learning and simulation (KnoCS)

Successful operation of scientific simulations and machine learning requires knowledge about the context of these processes. The explication of such knowledge may constitute methodological progress benefiting data-integrated simulation science. This project aims to acquire and represent formal knowledge that contextualizes the individual and the integrated processes of machine learning and simulation. Such context knowledge formalizes assumptions about models, expressing which models might be preferred or which models might be applicable in which context.

Specifically, we aim to answer the following research questions:

1. How does simulation knowledge contribute to the machine learning of surrogate models?
2. How can process knowledge help interpret parameters acquired from machine learning?
3. What are the challenges and opportunities for using explicit knowledge in machine learning and simulation in general?