Workshop Schedule

All paper talks are 15 minutes plus 5 minutes for questions.

The SC agenda is here:
https://sc21.hubb.me/fe/schedule-builder/sessions/880249

Time SessionEvent
8:30Invited SpeakerGeorgia Tourassi (ORNL)
Title: Leadership Computing in Support of Cross-Facility Workflows
The DOE Leadership Computing Facilities (LCFs) support extreme-scale computational and data campaigns that are typically one or two orders of magnitude larger than what might be feasible in typical institutional clusters. These campaigns at the extreme end of computing lend themselves to high-resolution exploration of natural systems in which theoretical models are used to delve into increasingly complex system dynamics. In the last decade, large-scale data analysis phases have become a vital part of scientific modeling and simulation-driven computational campaigns. The data analysis at scale is needed for large-scale AI training, and crucially, for processing data from observational facilities generating higher resolution voluminous data. Processing data from observational science facilities offers the promise of executing cross-facility campaigns more efficiently, and steering the physical experiment in real-time through co-executing computation. I will discuss emerging design patterns for intra- and cross-facility workflows, and raise important challenges that are ahead of us in making such computational and data environments a reality.
9:00Papers 1: Infrastructure
Chair: Debbie Bard (LBL)
ALS Share, a Lightweight Data Sharing Service for a Synchrotron Radiation Facility
by Bear, Parkinson, Gerhardt, Sibony, and Enders
Designing a Streaming Data Coalescing Architecture for Scientific Detector ASICs with Variable Data Velocity
by Strempfer, Yoshii, Hammer, Miceli, and Bycul
Bridging Data Center AI Systems with Edge Computing for Actionable Information Retrieval
by Liu, Ali, Kenesei, Miceli, Sharma, Schwarz, Trujillo, Yoo, Coffee, Layad, Herbst, Thayer, Yoon, and Foster
10:00Break
10:30Papers 2: Algorithms
Chair: Christine Sweeney (LANL)
High-Performance Hybrid-Global-Deflated-Local Optimization with Applications to Active Learning
by Noack
Adversarial Attacks against AI-driven Experimental Peptide Design Workflows
by Ramanathan and Jha
Optimizing High-Throughput Capabilities by Leveraging Reinforcement Learning Methods with the Bluesky Suite
by Olds, Allan, Caswell, Lynch, Maffettone and Campbell
11:30Panel
Chair: Eli Dart (ESnet)
Panelists: Thomas Caswell (BNL), Kurt H Maier (PNNL), Mark Neubauer (UIUC)
11:55AwardsXLOOP Chairs
Best paper
Best presentation

XLOOP Presentation Poll: https://forms.gle/NBTvWAUYNpdpK5nt5