A3D3 Kick-off meeting

8 Nov 2021, 11:00 → 10 Nov 2021, 13:10 US/Pacific

The National Science Foundation (NSF), under the Harnessing the Data Revolution (HDR) program, is providing funding to establish the Accelerated AI Algorithms for Data-Driven Discovery (A3D3) Institute, a multi-disciplinary and geographically distributed entity with the primary mission to lead a paradigm shift in the application of real-time artificial intelligence (AI) at scale to advance scientific knowledge and accelerate discovery. 

The A3D3 Institute was funded on 1 October 2021, and is ramping up its activities. This kickoff workshop brings together members and collaborators of the A3D3 to develop the activities which will be pursued in the five years of the Institute. The goal, motivation, deliverables, milestones, and timeline will be identified for each activity. We should leave the workshop with a comprehensive Project Execution Plan. This plan will be used to develop the Management Plan requested by the NSF due mid of December. 

Monday, 8 November

1 Introduction

Speaker: Shih-Chieh Hsu (University of Washington Seattle (US))

gslide

2 Logo and Website

Speakers: Miaoyuan Liu (Purdue University (US)), Philip Coleman Harris (Massachusetts Inst. of Technology (US))

logo

website

Lightning talks

3 HEP

Speaker: Miaoyuan Liu (Purdue University (US))

a3d3_kickoff_hep.pdf

4 MMA

Speaker: Michael Coughlin (University of Minnesota)

MMA HDR Planning.pdf

5 Neuros

Speaker: Amy Orsborn

kickoffMeeting_neuroIntro_v211107.pdf

6 HAC

Speakers: Pan Li, Pan Li

Lightening-talk-HAC.pdf

7 Targeted & Heterogenous Systems

Speakers: Eli Shlizerman, Javier Mauricio Duarte (Univ. of California San Diego (US))

A3D3_TargetedSystems_7Nov2021.pdf

8 Engagement

Speakers: Mark Neubauer (Univ. Illinois at Urbana Champaign (US)), Mark Stephen Neubauer (Univ. Illinois at Urbana-Champaign)

A3D3_Engagement_kickoff.pdf

9 Education

Speakers: Matthew Graham (California Institute of Technology), Matthew Graham

10 Equity and Career

Speakers: Javier Mauricio Duarte (Univ. of California San Diego (US)), Miaoyuan Liu (Purdue University (US))

A3D3 Kickoff_ Equity & Career.pdf

HEP

Convener: Miaoyuan Liu (Purdue University (US))

11 Graph neural networks in HEP

10 mins talk + 5 mins discussion

Speaker: Yongbin Feng (Fermi National Accelerator Lab. (US))

GNN_Overview_A3D3.pdf

12 Anomaly detection in HEP

10 mins talk + 5 mins discussion

Speaker: Sang Eon Park (Massachusetts Inst. of Technology (US))

A3D3 Final 1.pdf

13 SPVCNN applications in HEP

10 mins talk + 5 mins discussion

Speaker: Alexander Joseph Schuy (University of Washington (US))

SPVCNN slides

MMA

Convener: Michael Coughlin (University of Minnesota)

Neuroscience

Convener: Amy Orsborn

14 intro to neuro data

overview of signal types, properties

Speaker: Prof. Maria Dadarlat (Purdue)

15 auto-encoders for neural signal prediction

Speaker: MICHAEL NOLAN

Electrocorticograph Signal Modeling with RNN Autoencoders.pptm

16 approaches to analyze behavior

Speaker: JINGYUAN LI

Approach to Analyze Behavior Sparse Semi-Supervised Action Recognition with Active Learning.pdf

17 discussion time

HAC

Convener: Pan Li

18 Intro

HAC (Hardware and Algorithm co-development) is one main research direction of A3D3, where the goal is to study the state-of-the-art AI hardware and algorithms for scientific applications. In this session, we have four talks that include AI algorithm/hardware design for physical applications, and efficient compilation tools for AI hardware automatic design.

Speaker: Pan Li (Purdue)

19 Point cloud methods for science

Speaker: Zhijian Liu (MIT)

20211108 - A3D3.pdf

20 GNN-for-Science

Speaker: Siqi Miao (Purdue)

tau3mu-siqi[17973].pdf

21 ScaleHLS

Speaker: Hanchen Ye (UIUC)

A3D3_ScaleHLS_Slides_Hanchen[17518].pdf

22 HLS4ML

Speaker: Dylan Rankin (MIT)

a3d3_hac_dsr_01Nov21 (1).pdf

Tuesday, 9 November

Targeted and heterogeneous systems

Conveners: Eli Shlizerman, Javier Mauricio Duarte (Univ. of California San Diego (US))

23 Intro

Targeted systems: hardware-based, edge devices, on-sensor, including FPGA, ASIC, or other technology used for monitoring, instrument control, data selection, or data processing often in real-time or with very low latency.

Heterogenous systems: CPUs + coprocessors, high-throughput data processing, may include HPCs, cloud resources, asynchronous communication, as-a-service.

This section will give a brief overview of the different systems we target in HEP, neuroscience, and MMA with a look toward how we can develop common tools and techniques. Each talk will address:

1. What the targeted/heterogeneous systems look like (hardware specifications, constraints, etc.)
2. How we currently "program" these targeted systems (including the use of AI algorithms and tool flows like hls4ml)
3. Thoughts for improving workflows, infrastructure, which can be developed in common

Speakers: Eli Shlizerman, Javier Mauricio Duarte (Univ. of California San Diego (US))

24 HEP: Level-1 trigger (FPGA) + On-detector readout (ASIC)

Speaker: Daniel Diaz (Univ. of California San Diego (US))

A3D3_HEP-L1Trigger.pdf

25 HEP: Services for Optimized Network Inference on Coprocessors (SONIC)

Speaker: Jeffrey Krupa (Massachusetts Institute of Technology)

a3d3_krupa.pdf

26 MMA: Gravitational Wave Inference as a Service

Speakers: Alec Gunny, Alec Gunny (Massachusetts Institute of Technology)

A3D3 Kickoff Presentation - LIGO.pdf

27 Neuroscience: Sleep Spindles as a Driver of Low Latency, Low Power ML

Speaker: Xiaohan Liu

Sleep Spindles as a Driver of Low Latency, Low Power ML in HLS4ML & TinyML.pdf

Diversity, Equity, Inclusion, & Career

Conveners: Javier Mauricio Duarte (Univ. of California San Diego (US)), Miaoyuan Liu (Purdue University (US))

28 Post-bacc Rubric Discussion

Speaker: Javier Mauricio Duarte (Univ. of California San Diego (US))

A3D3_HolisticReview.pdf

Proposed rubric

29 Post-bacc Ad

postbac ad draft

30 Code of Conduct

CodeOfConduct

31 Postdoc Mentoring

postdoc mentoring

Education

Conveners: Matthew Graham (California Institute of Technology), Matthew Graham

gdoc

Engagement

Conveners: Mark Neubauer (Univ. Illinois at Urbana Champaign (US)), Mark Stephen Neubauer (Univ. Illinois at Urbana-Champaign)

32 Engagement with other Communities/Projects and CEL Discussion

Speaker: Mark Neubauer (Univ. Illinois at Urbana Champaign (US))

Session Notes

33 Resources: HAL at NCSA's Innovative Systems Laboratory

Speaker: Volodymyr Kindratenko (University of Illinois at Urbana-Champaign)

caii_dl_tech.pdf

34 Resources: XACC Cluster at Illinois Coordinated Sciences Laboratory

Speaker: Deming Chen (University of Illinois at Urbana-Champaign)

UIUC_XACC.pdf

35 Resources: Cloud

Speaker: Philip Coleman Harris (Massachusetts Inst. of Technology (US))

PCH_Cloud_09_11.pdf

Wednesday, 10 November

Summary

Convener: Philip Coleman Harris (Massachusetts Inst. of Technology (US))

36 HAC

Speaker: Pan Li (Purdue University)

HAC-summary.pdf

37 HEP

Speaker: Miaoyuan Liu (Purdue University (US))

38 MMA

Speaker: Michael Coughlin (University of Minnesota)

MMA HDR Planning.pdf

39 Neuroscience

Speaker: Amy Orsborn

Neuroscience in A3D3_Summary and Discussion.pdf

40 Targeted & Heterogeneous systems

Speakers: Eli Shlizerman, Javier Mauricio Duarte (Univ. of California San Diego (US))

A3D3_TargetedSystems_10Nov2021.pdf

41 Diversity, Equality and Inclusive

Speakers: Javier Mauricio Duarte (Univ. of California San Diego (US)), Miaoyuan Liu (Purdue University (US))

42 Education

Speakers: Matthew Graham (California Institute of Technology), Matthew Graham

43 Engagement

Speakers: Mark Neubauer (Univ. Illinois at Urbana Champaign (US)), Mark Stephen Neubauer (Univ. Illinois at Urbana-Champaign)

A3D3_Engagement_kickoff_meeting_summary.pdf

44 Final Remarks

Speaker: Shih-Chieh Hsu (University of Washington Seattle (US))

gslide