2024-07-10 –, Else (1.3)
Structural identifiability is a property of a dynamical model that determines if the model parameters can be inferred uniquely in the absence of noise. StructuralIdentifiability.jl allows to assess the identifiability of dynamical models described by ODEs. In this talk, we will present new functionality in StructuralIdentifiability.jl, which allows one to find all identifiable combinations in the model under consideration and automatically reparametrize the model to make it fully identifiable.
As a result of recent work, new functionality has been implemented in StructuralIdentifiability.jl. We would like to introduce this functionality in our talk. This new functionality allows the user to find, in a matter of seconds, all rational functions of parameters and states that are identifiable in the model of consideration and, possibly, reparametrize it. To the best of our knowledge, other open-source state of the art software cannot do that.
Identifiability analysis and reparametrization is an important step of model design, so we expect that the topic of the talk would be of interest to practitioners in modeling and control.
During the presentation, we will try to avoid the technicalities, and instead focus on simple but meaningful examples. We plan to include a short demo to demonstrate the package capabilities.
For further details, one can consult the relevant tutorial pages from the package documentation:
https://docs.sciml.ai/StructuralIdentifiability/v0.5.2/tutorials/identifiable_functions/
https://docs.sciml.ai/StructuralIdentifiability/v0.5.2/tutorials/reparametrization/
This work is a joint result by Alexander Demin, Gleb Pogudin, and Chris Rackauckas.
I am a senior at HSE University in Russia. I major in mathematics and computer science.