NPPS Leadership Team

[Brett Viren <bv AT bnl.gov>]

Aschenauer Elke-Caroline via Phys-npps-mgmt-l
<phys-npps-mgmt-l AT lists.bnl.gov> writes:

> I like Brett’s argument maybe this should be made more explicit on the 
> slides.

And the "why BNL?" question is right on target.

> Becuase indeed AI./ML is something which is done since quite some time it 
> was not so called out.
> But the BNL HEP expertise about the systems/analysis it is applied to is 
> unique.
>
> I would drive this point home.

I think we can claim with a straight face that BNL brings unique and
substantial "domain knowledge" needed for application of AI/ML to LArTPC
problems and in ways that surpass other institutions.

The strongest argument I see is that BNL has demonstrated the world's
best LArTPC software in three areas: detector simulation, signal
processing and reconstruction.  All three are critical for effective
application of AI/ML in this domain.  We know what works and we've
implemented that.  And in doing that we have learned where AI/ML likely
can be fruitful, even outright needed.  And, more importantly where
applying it is likely a waste of time.

BNL is also unique in having both the Instrumentation Division for their
LArTPC expertise (they invented much of it so are clearly in a unique
position!) and CSI for their AI/ML expertise (among other things).  Of
course, CSI involvement can benefit all our AI/ML efforts, not just
those in the domain of LArTPC.


-Brett.

PS: The "best reco" claim may seem like bravado but it will be
demonstrated as part of new MicroBooNE results which are expected out
real soon now.  The "best sim and sigproc" claim has been established
for a while.

Attachment: signature.asc
Description: PGP signature