Reconstruction vs Identification

[ChrisVer]

This is a fast-question/need of confirmation concerning the two: Reconstruction and Identification.
I am trying to "separate" what they mean.
For example when we reconstruct a jet, we apply several selection criteria that the object "jet" should satisfy. Is that correct? Then any object that satisfies those criteria is reconstructed as a jet and used as one afterwards. My confusion comes here because this implies an identification; the object is seen as a jet afterwards. Or I should better say reco-jet.
Identification on the other hand is also using other selection criteria + some variables fed in a BDT (or LLH) that are used to identify the jet as a fake (maybe from a tau) or not?

If the above is "correct" (if it's wrong then I have misunderstood something crucial), I don't see the reason for applying the two and not just going for ID alone. it's like going from city A (bulk events) to B("known" event type and kinematics) by passing first from another city C (reconstruction), rather than taking the straight way to B.
Thanks

 

[mfb]

For example when we reconstruct a jet, we apply several selection criteria that the object "jet" should satisfy. Is that correct?

Right.

Reconstruction and identification are not always clearly separated. Experimentally, you often have a two-step process. Reconstrution gives you "here, this could be a jet" and identification then says "okay, this jet candidate is probably from the right primary vertex and satisfies some other quality criteria".
Generic jets don't have strong identification criteria, because nearly everything with energy deposit in the calorimeter is called a jet. The distinction is clearer for everything else.

Identification can use a BDT but does not have to.

 

[ChrisVer]

Identification can use a BDT but does not have to.

hmm... I can't see that...
Wouldn't the ID come with a probability that the particle is what it's IDed with? I thought that BDT does that immediately by separating bkg to signal (and result to loose,medium or tight IDs). Is there an alternative way?

 

[ChrisVer]

ah... I think instead of feeding the variables in a BDT you could apply cuts on them...

 

[mfb]

ah... I think instead of feeding the variables in a BDT you could apply cuts on them...

Right.
And even with a BDT, the output is often used as "yes/no" - a cut on the BDT output.