Is it possible to run a non-tv filter for a linear system? I have a script running a filter with great accuracy that is attempting to model a constant variance scenario, however the state means are not similar to results from other filters (filterpy/pykalman) and exhibit odd behaviors (assuming this is modeling nonlinear relationship). Seems as though I cannot seem to properly adapt the settings for the constant case, which embarrassingly sounds trivial. Is there a simple solution or any examples I could be pointed at?
Edit: Got it working. This implementation is blazingly fast and very accurate. Great work
So glad this is useful for you! We'll be upgrading to a JAX version of the library at some point. It should be just as fast but amenable to automatic differentiation :)
Is it possible to run a non-tv filter for a linear system? I have a script running a filter with great accuracy that is attempting to model a constant variance scenario, however the state means are not similar to results from other filters (filterpy/pykalman) and exhibit odd behaviors (assuming this is modeling nonlinear relationship). Seems as though I cannot seem to properly adapt the settings for the constant case, which embarrassingly sounds trivial. Is there a simple solution or any examples I could be pointed at?
Edit: Got it working. This implementation is blazingly fast and very accurate. Great work