Conversation
…. Also, improvements in refinement of already found tracks (e.g. dropping bad hits). Negative covariance still does happen, so more improvements are in the works. At least these changes prevent those tracks from becoming completely crazy.
…nal formula. The new formula clearly wasn't working.
|
Hi All, Can people help reviewing this PR and approving it. From my side looks fine and I'd like to test this asap. |
|
I'm trying to figure out how best to test this. Perhaps we can discuss it at either the analysis or recon meeting on 4/13. |
|
Ok for recon meeting tomorrow then =) |
| Cinv.unsafe_set(i,j,1.0/Cinv.unsafe_get(i,j)); | ||
| } else { | ||
| Cinv.unsafe_set(i, j, 0.); | ||
| SquareMatrix invrs = KalTrack.mToS(snP.helix.C).fastInvert(); |
There was a problem hiding this comment.
For future, this can easily be turned into a stream with a lambda instead. Don't think it's needed now, but wanted to note it.
There was a problem hiding this comment.
This applies to most for loops ...
|
At some point, we should move to using a eigen or some other math library ... again, future cleanup. |
cbravo135
left a comment
There was a problem hiding this comment.
It appears distribution/pom.xml has not been updated yet to the 5.1 snapshot. Since git says this can be automerged, this small issue will automatically be dealt with. These changes are isolated to details of the KF tracking, which Robert explained during the reconstruction meeting. Testing will happen before we decide to run on any serious amount of data and additional changes can still be made going forward.
|
I merged it in. Since KF tracking is still in dev, more iterations will be done here for test this properly in the future |
No description provided.