SMC 2020 : Smoky Mountains Computational Sciences & Engineering Conference

This year, the program committee will accept vision papers that include the author’s perspective on the most important directions for research, development, production and experiences, and needs for investment in the specific areas identified in the following five sessions:

Session 1. Computational Applications: Converged HPC and Artificial Intelligence (AI)

Session chairs – Bronson Messer and Steven Hamilton, ORNL

This session will address applications that embrace data-driven and first-principle methods, focusing on converging AI methods and approaches with high-performance modeling and simulation applications. Topics will include experiences, algorithms, and numerical methods that will play an important role in this area. Participants will discuss how simulation can be used to train AI models and integrate them to work with simulation applications while quantifying errors.

Session 2. System Software: Data Infrastructure and Life Cycle

Session chairs – Sudharshan Vazhkudai and Amy Rose, ORNL

In this session, participants will consider the scientific data life cycle from collection to archive, including all the aspects in between and the infrastructure needed to support it. The group will cover techniques and system designs needed to securely publish, curate, stage, store, reduce, and compress data. Also relevant are techniques to annotate the data with metadata and automatically extract information from datasets that will aid with the scalable search and discovery of mountains of data.

Session 3. Experimental/Observational Applications: Use Cases That Drive Requirements for AI and HPC Convergence

Session chairs – Kate Evans and Vincent Paquit, ORNL

Participants will discuss ways to use multiple federated scientific instruments with data sets and large-scale compute capabilities, including sensors, actuators, instruments for HPC systems, data stores, and other network-connected devices. Some of the AI and HPC workloads are being pushed to the edge (closer to the instruments) while large-scale simulations are scheduled on HPC systems with large capacities. This session will focus on use cases that require multiple scientific instruments, emphasizing use cases that combine AI and HPC with edge computing. Priority areas of interest include, but are not limited to:

• Use cases that require multiple scientific instruments,
• Use cases that combine AI and HPC with edge computing,
• Examples that show how to interface with the users (of data, output..)
• Examples that demonstrate a diversity of connectivity across instruments, HPC, data store,
• Novel methods within a larger use case that show improvements one facet, say edge compute, storage, transfer etc. that enhance current practice,
• An overview of the state of convergence of data, compute, and instruments in an application area, and
• Provocative ideas on how to revolutionize the convergence of instruments, data, and compute at the edge and HPC.

Session 4. Deploying Computation: On the Road to a Converged Ecosystem

Session chairs – Gina Tourassi and Arjun Shankar, ORNL

Topics will include industry experience and plans for deploying the hardware and software infrastructure needed to support applications used for AI methodologies and simulation to deploy next-generation HPC and data science systems. This session will focus on how emerging technologies can be co-designed to support compute and data workflows at scale.

* Session 5. Scientific Data Challenges: Data Sponsors

Session chair – Suzanne Parete-Koon, ORNL

SMC2020 provides scientists with an opportunity to become scientific data sponsors and describe challenges for eminent data sets at ORNL. These data sets will be used for the SMC Data Challenge (SMCDC2020) competition (https://smc-datachallenge.ornl.gov). These data sets come from scientific simulations and instruments in physical and chemical sciences, electron microscopy, bioinformatics, neutron sources, urban development, and other areas. The goal of this session is to provide and describe a significant data set, then formulate three to five challenge questions associated with the data set in a paper. The challenge questions for each data set will cover multiple difficulty levels. The first question in each challenge should be suitable for a novice, with each subsequent question increasing in difficulty and the series of questions ending with an advanced/expert level challenge question. These challenges are intended to draw scientists and researchers at the beginning stages of incorporating data analytics into their workflow, as well as data analytics experts interested in applying novel data analytics techniques to data sets of national importance.

* Data sponsor papers are invited papers and don’t need to submit an abstract