Closed sdhiscocks closed 4 months ago
All modified and coverable lines are covered by tests :white_check_mark:
Comparison is base (
0a9ebe2
) 93.24% compared to head (2b3d3f1
) 93.24%.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
@narykov Does this resolve #941?
@narykov Does this resolve #941?
Apologies I didn't get back to this earlier. I can confirm this has resolved the issue. Thank you!
@narykov Does this resolve #941?
Apologies I didn't get back to this earlier. I can confirm this has resolved the issue. Thank you!
Perhaps, the only additional comment I have is that in my case each prediction in the track is accompanied by its own transition_model
, so I find it puzzling that I am forced to specify a single transition model when calling the smoother. As a consequence, I had to specify some dummy model that the smoother ends up never using.
Apologies I didn't get back to this earlier. I can confirm this has resolved the issue. Thank you!
Thanks for confirming.
Perhaps, the only additional comment I have is that in my case each prediction in the track is accompanied by its own
transition_model
, so I find it puzzling that I am forced to specify a single transition model when calling the smoother. As a consequence, I had to specify some dummy model that the smoother ends up never using.
So hopefully setting it to None
should work okay? But maybe we should make it default None
so it's optional?
So hopefully setting it to
None
should work okay? But maybe we should make it defaultNone
so it's optional?
Setting transition_model=None
works for me, just it hasn't crossed my mind that there is that possibility. And as for the second question, I am thinking that it is reasonable to have transition_model
explicitly specified as optional because:
The supplied transition_model
is primarily used as a fallback option, which won't be used unless the track updates lack their own models. Furthermore, one might mistakenly believe that it is the supplied model that is employed for processing, whereas another model is actually implemented inside the smoother.
I believe that one could obtain a Track
object without providing a reference to any transition_model
when the updates are constructed using the from_state
method. However, this is by no means a common way of using Stone Soup.
Thanks!
The supplied transition_model is primarily used as a fallback option, which won't be used unless the track updates lack their own models. Furthermore, one might mistakenly believe that it is the supplied model that is employed for processing, whereas another model is actually implemented inside the smoother.
Perhaps, if transition_model
is specified explicitly in the definition of smoother it should be enforced by the smoother? As this appears to me as the expected behaviour.
Fixes #941