search

acse-jm122 / torchda

Use Deep Learning in Data Assimilation
GNU General Public License v3.0
21 stars 4 forks source link

data assimilation for system model #2

Open Luoxiaoyu828 opened 1 month ago

Luoxiaoyu828 commented 1 month ago

I would like to know, for a complex system model that, when given model parameters and initial state parameters, can iteratively run and eventually produce stable output values. When the state parameters change, after a certain number of iterations, the model will once again output stable parameters. In such a scenario, where noise is added to the model output to simulate observational data, and while the system model can be called, other model state transitions and observation functions are unknown, could data assimilation be performed using your TorchDA framework? If so, what data would need to be prepared?

acse-jm122 commented 1 month ago

@Luoxiaoyu828 Sorry, I don't quite fully understand your concern about performing this task on TorchDA. It seems like you are dealing with a normal input-model-output end to end case, but it would be obvious to you that using normal framework such as PyTorch or TensorFlow can solve it. I am wondering what's the "a certain number of iterations" referring to? Is that referring to performing optimization steps on model parameters or passing representation of state parameters through the model repeatedly?