David Broman - Domain-specific modeling: equation-based languages and probabilistic programming

Models are extensively used by both scientists and engineers, but for different reasons. For scientists, working in natural sciences, the main motivation for constructing models is to understand the system being modeled. By contrast, an engineer typically makes use of models when creating systems. These models can be used in many types of analysis, such as formal verification, statistical analysis, or simulation. This talk focuses on languages for modelling, in particular on languages that are domain specific. I will discuss two specific kinds of domain-specific languages: (i) equation-based languages for modelling and simulation of cyber-physical systems, and (ii) probabilistic programming languages for performing Bayesian inference. Although there are several benefits with such domain-specific modeling languages, the cost of developing efficient environments and compilers can be very high. Specifically, people with strong domain knowledge (for instance in biology or mechanical engineering) are seldom compiler experts. Likewise, compiler writers rarely have strong domain knowledge, especially not in multiple scientific or engineering domains. I will discuss our ongoing effort of developing a general compiler infrastructure for efficient domain-specific language construction, as well as various applications domains, including phyogenetics, health care, and mechanical engineering.

Видео David Broman - Domain-specific modeling: equation-based languages and probabilistic programming канала Digital Futures: Research Hub for Digitalization