GSI Forum: Tracking » cov[0][0] = 0 in Kalman.cxx

Home » PANDA » PandaRoot » Tracking » cov[0][0] = 0 in Kalman.cxx

Show: Today's Messages :: Polls :: Message Navigator

cov[0][0] = 0 in Kalman.cxx [message #6980]

Thu, 19 June 2008 11:24

Susanna Costanza
Messages: 33
Registered: January 2008
Location: Pavia

continuous participant

From: *pv.infn.it

Hi Christian!
I see that in one of the latest revision you have added an exception if cov[0][0] is 0 in Kalman.cxx.
Can you tell me how these cases are handled? ...because I'm not so keen on exceptions!
Do you throw away the whole track or just that hit, skipping to the next hit and returning the previous rep?

Ciao,
Susanna

Report message to a moderator

Re: cov[0][0] = 0 in Kalman.cxx [message #6981 is a reply to message #6980]

Thu, 19 June 2008 11:29

Anonymous Poster

From: *pool.mediaWays.net

Hi Susanna,

I throw away the whole track in this case. You can see that the processHit is done inside a try-catch block. If the exception happens the statusFlag goes to 1 and the track is aborted.

The cov[0][0]==0. should never happen. However it does sometimes. If it doesnt anymore that would be even better Smile

Christian

Report message to a moderator

Re: cov[0][0] = 0 in Kalman.cxx [message #6982 is a reply to message #6981]

Thu, 19 June 2008 15:03

Susanna Costanza
Messages: 33
Registered: January 2008
Location: Pavia

continuous participant

From: *pv.infn.it

Thank you for your explanation about the use of exceptions!

I did some tests... What I see is that throwing away all the events in which cov[0][0] is null results in a considerable loss of efficiency.
Instead, if you skip just that particular hit in which the covariance matrix has that problem, GEANE propagates to the next hit and sometimes the track returns to the right path. The resulting Kalman rep is quite good and the efficiency is higher.

Ok, I agree that cov[0][0] should never be 0 and we will take a look in order to understand why there is this strange behaviour for some events...
In the meantime, do you think it's a so bad idea to "convert" the exception into a return, in order to save a bigger number of tracks?

Ciao,
Susanna

Report message to a moderator

Re: cov[0][0] = 0 in Kalman.cxx [message #6983 is a reply to message #6982]

Thu, 19 June 2008 15:15

Anonymous Poster

From: *e18.physik.tu-muenchen.de

Hi,

no I dont think that it is such a bad idea and I can change this. I will do it in the next days. Right now I am a little backed up...

However, it would be really nice to understand what is going on with the cov[0][0]=0.

Bye, Christian

Report message to a moderator

Previous Topic:	GEANE backward propagation
Next Topic:	New Features in GENFIT

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Sun Nov 24 09:22:42 CET 2024

Total time taken to generate the page: 0.00843 seconds