A3D3 Steering Board

US/Pacific
Philip Coleman Harris (Massachusetts Inst. of Technology (US)), Shih-Chieh Hsu (University of Washington Seattle (US))
Description

Steering Board is composed of representatives from the A3D3 stakeholders and act as an advocate for initiatives and projects across the institute. Set the strategic direction of projects and provide support, guidance and oversight of progress.

Steering Board members are elected by the A3D3 senior personnel annually. The committee meet with Executive Board on the last Friday of each quarter. The feedback from the Steering Board will be included in the A3D3 quarterly report.

Steering Board members:

  • Internal Members
    • Song Han (HAC)
    • Mark Neubauer (HEP)
    • Kate Scholberg (MMA)
    • Maria Makin (Neuro)
    • TBA (Young scientist)
  • External Members
    • Nhan Tran (HEP)
    • V. Ashley Villar (MMA) 
Videoconference
A3D3 Executive Board
Zoom Meeting ID
65963516231
Host
Shih-Chieh Hsu
Useful links
Join via phone
Zoom URL
  • MMA

    • How are the institutes connected?

    • How three areas (GW, Neutrino, Optical) converged?

    • Are all parallel technologies?

    • What is the goal in the next 3months?

      • Establish data cleaning pipeline for GW

  • HAC

    • How to manage three different teams?

      • There seems to be a lot of work on pure ML

      • Is there a way to mitigate this? 

    • Projects are loosely connected

    • Extending A3D3 to external associated members?

      • ML group can use more experts outside A3D3

    • How to connect pyLOG to HLS4ML

      • Cross-meeting presentations  in HLS4ML meeting and vice sa

    • Is this work being submitted to interdisciplinary journals/conferences? 

      • For example, there was a neurips/icml submission. Were these submitted to the main conferences or specialized workshops? Does the team have a general plan on publication goals?

  • HEP

    • What overlap among various groups and projects?

  • Neuros

    • What is collaboration platform?

      • Maria and Amy are generating broader experiment data and each has customized system so far. Ideas is to develop OPEN DATA set as new benchmark dataset initiated by A3D3

      • Just a comment as someone who is very much outside of the field: an open source data publication could be very high impact!! Astro creates Kaggle competitions to solve some of our problems, which are very popular (Ashley)

    • What is the hardware system? (are they very diverse, many vendors?)

    • Workshop in pure neuroscience building (30K people…wow. Society of Neuroscience)

  • Targeted system

    • PyLog/ScaleHLS integration to HLS4ML

      • Biweekly meeting with Vladimor+HLS4ML team and UIUC team

      • Identify cross projects to bring people together

  • Community Engagement 

    • Hardware/Industry, have you engaged with the Xiliinx(AMD) university program? 

      • Yes!

    • How do you engage domains beyond HEP?
      In particular, Society of Neuroscience, Medical Physics society

    • What do we want to accomplish in the next 3 months? 

      • Aim for an HDR leadership meeting

      • Work on assembling and distributing FAIR datasets

      • Increasing industry engagement

    •  
  • Education

    • Building A3D3 FAIR dataset for education

    • There are a lot of educational efforts going on, would it be possible to centralize this effort, and make a common effort

      • Common tutorials

      • Common Datasets

      • Sharing of problems

    • How to bring in ML and ECE people for education 

      • I just say that maybe Song and Deming might also have a course around this too — I know that Luca Carloni at Columbia has a research class that advertises science projects to (Nhan)

    • Target summer schools

      • Detector schools (grad/undergrad), AI schools, Radio Astronomy schools

    • Chat comments

    • We have an event building telescopes in a local elementary school here April 21 (Michael)

    • I have a high school student who currently has a PYNQ-Z2 is looking at adapting our hls4ml tutorial for high school students, but very early days (Javier)

  • Equity and Diversity

    • 14 in comparison to 100 applicants in FNAL equivalent AI Postbac program. We have quite 

    • Recruiting platform: AJO, CMS mail list, University mail list, APS education program, …, add a complete ist into the quarter report.

      • Grace Hopper is really good (used by FNAL)

    • Can you consider advertising at more society booths

    • Can you clarify whether or not these students will receive advanced payment to help with moving costs? This would likely be extremely beneficial for the audience you’re targeting (Yes. That can be done)

    • What mentor structures are set up to ensure they will be set up for grad school applications?

  • Summary

    • Nhan:

      • CS expertise, multidisciplinary conversation, 

      • How can we consolidate the efforts of Song, Deming and HLS4ML

    • Ashley: OpenLab, ML methodology?  

      • ML group seems a bit overburdened

        • Coming up with a way to address this would help

      • For next meeting, can we can make a clear list of what was done in the quarter

    • Maria: 

      • Work towards standardization of algorithms, datasets

        • Develop the toolkit to go cross domain

    • Kate:

      • Distribute Steering Board questions 2 weeks ahead 

    • Phil:

      • Associate member/unpaid member 

There are minutes attached to this event. Show them.
    • 15:00 15:15
      Introduction 15m
      Speakers: Philip Coleman Harris (Massachusetts Inst. of Technology (US)), Shih-Chieh Hsu (University of Washington Seattle (US))
    • 15:20 17:20
      Vision and Progress
    • 16:50 17:10
      Steering Board Feedback & Discussion 20m