US ATLAS HL-LHC computing discussion

[Daniel Murnane <dtmurnane AT lbl.gov>]

Slides attached!

Daniel

On Mon, Mar 2, 2020 at 1:54 PM <kaminskaya AT google.com> wrote:

This event has been changed with this note:
"Look forward to seeing everyone in person or on Google Video Call tomorrow! I just added a brief agenda to the calendar invite. "

Atlas ML Track: Tech meetings at Google, Sunnyvale, CA

When

Tue Mar 3, 2020 10am – 12pm Pacific Time - Los Angeles

Where

SVL-MP2-8-Jannu (13) [GVC] (map)

Joining info

Join Hangouts Meet

meet.google.com/vqf-buju-atx

Join by phone

+1 319-343-8105 (PIN: 149094)

More phone numbers

Calendar

dtmurnane AT lbl.gov

Who

•	kaminskaya AT google.com - organizer
•	dtmurnane AT lbl.gov
•	heathg AT gmail.com
•	pcalafiura AT lbl.gov
•	jarekk AT google.com
•	njchoma AT lbl.gov
•	sfarrell AT lbl.gov
•	xju AT lbl.gov
•	charles.g.leggett AT gmail.com
•	markmims AT google.com
•	Ilija Vukotic
•	usatlas-hllhc-computing-l AT lists.bnl.gov
•	Malik,Abid
•	Lin, Meifeng
•	Farbin, Amir
•	kaushikde.uta AT gmail.com - optional
•	mileseu AT google.com - optional
•	kajameson AT google.com - optional
•	Usman Qureshi - optional
•	wenaus AT gmail.com - optional
•	Karan Bhatia - optional
•	aak AT bnl.gov - optional

more details »

Changed:
Agenda: 

10:00 - 10:30 am  Trace reconstruction problem background 
10:30 - 11:30 am   Discussion: Google Cloud & where we can help your ML efforts  
11:30 - 11:45  Action Items & Next Steps
--------------------------------------------------------------------------------------------------------------------
Hi All,

Please hold this date, March 3rd, 10 am to 12 pm Pacific Time for an Atlas ML Track Meeting at Google. You can attend in person or join the video call.

The goal of the meeting is to discuss Paolo's team tracking algorithm and see what improvements can be made to it.

Please let me know if the proposed date/time works for you and I will start planning the meeting.

Ema
------------------------------------------------------------------------------------------------------------------------------------------
Background Information from Paolo:


The Kaggle challenge and dataset are at
https://www.kaggle.com/c/trackml-particle-identification 
The page contains several notebooks that help understand the problem and the language.

We currently have a prototype algorithm that uses Graph Neural Network to solve the problem
https://github.com/exatrkx/exatrkx-neurips19/tree/master/gnn-tracking  


The GitHub also has a link to our neurips paper 

https://ml4physicalsciences.github.io/files/NeurIPS_ML4PS_2019_83.pdf


We have achieved excellent results in deciding which hit doublets (segments) are likely to belong to a track. We do not have a good architecture to solve the track following step of the algorithm: given a variable-length list of "good" segments, produce a variable-length list of likely track candidates. Currently, we are doing this with some simple heuristic that does not perform as well as the doublet classifier.

Going (dtmurnane AT lbl.gov)?   Yes - Maybe - No    more options »

Invitation from Google Calendar

You are receiving this email at the account dtmurnane AT lbl.gov because you are subscribed for updated invitations on calendar dtmurnane AT lbl.gov.

To stop receiving these emails, please log in to https://www.google.com/calendar/ and change your notification settings for this calendar.

Forwarding this invitation could allow any recipient to send a response to the organizer and be added to the guest list, or invite others regardless of their own invitation status, or to modify your RSVP. Learn More.

Attachment: ML for Particle Tracking.pdf
Description: Adobe PDF document