Three Argonne researchers will discuss their Laboratory-Directed Research and Development (LDRD) sponsored work at the LDRD Seminar Series presentation Tuesday, May 8, 2018, at 12:30 p.m. in the Building 203 Auditorium. All are welcome to attend.
Add to Your Calendar
Brian Craig
“Immersive 3-D Visualization of Military Complex Systems”
By Software Engineer Brian Craig (CFC)
Abstract
The U.S. military is constantly facing problems that are highly complex and adaptive. These problems are characterized by a high degree of interdependency, diverse perspectives of multiple stakeholders, emergent properties, and a high degree of uncertainty. To help understand these very complex problems, the military uses both quantitative and qualitative data to generate future scenarios, simulations and other analytical products. This LDRD explored virtually embedding users within the data to determine if we can improve our ability to explain, explore, and manipulate complex system properties, structures and dynamic emergent behaviors. A virtual environment was developed to place users inside of a display that geographically placed data that had been extensively pre-processed from a large body of free form text. The users are able to interact with the data, diving deeper into levels of hierarchy, as well as rearrange the data based on calculated relationships.
Biography
Brian A. Craig is a software engineer in the Integrated Analytics Group of the Systems Science Center of the Chemical and Fuel Cycle Technologies (Global Security Sciences). His work focuses primarily on the development of computer simulation models for various energy and supply-chain systems. He earned a B.S. and an M.S. in computer science from North Central College in Illinois.
Nicholas Jackson
“Enhancing Coarse-Grained Molecular Models using Machine Learning”
By Argonne Scholar Nicholas Jackson (MSD)
Abstract
The modeling of organic electronic materials involves the use of multiscale simulation techniques to accurately describe how molecular structure influences macroscopic performance. Critical to this multiscale effort is the link between coarse scale simulations that model mesoscopic length and time scales, and atomistic simulations that describe the behavior of electrons over nanoscopic length and time scales. In this talk, I describe how techniques from machine learning can be applied to enhance the interaction between models of different scales, dramatically accelerating the computational workflow, and associated understanding, of molecular simulations.
Biography
Nicholas Jackson received his B.A. in physics from Wesleyan University, Ph.D. in chemistry from Northwestern University and is currently a Maria Goeppert Mayer Fellow at Argonne supervised by Juan de Pablo (MSD) and Venkatram Vishwanath (ALCF). His research combines molecular modeling and machine learning to understand the thermomechanical and optoelectronic behavior of organic electronic and polyelectrolyte materials.
Min Si
“Designing Dynamic Communication Infrastructure for Scalable Irregular Parallel Computing”
By Argonne Scholar Min Si (MCS)
Abstract
The irregular computing model is being utilized by a number of emerging applications in domains such as chemistry, bioinfomatics and data analytics. Unlike the well-studied Bulk Synchronous Parallel (BSP) approach where the computation workload is distributed evenly across participating processors with synchronization at regular intervals, the irregular model focuses on a dynamic load balancing scheme and is usually characterized by unpredictable and fine-grained data access across processors and computing nodes. As the most widely used communication model over clusters and supercomputing systems, MPI defines the nonblocking message passing as well as the one-sided communication for the semantic support of irregular and asynchronous data access. The runtime infrastructure, however, still cannot take over such data movement efficiently.
In this talk, we first showcase the crucial performance bottlenecks existing in irregular computing through the case study of a large quantum chemistry application. We discover that the efficiency of irregular data movement is extremely restricted by the underlying resource assignment and scheduling for communication and computation. We then introduce a dynamic execution framework that is comprised of several runtime techniques and addresses these challenges for MPI applications.
Biography
Min Si is an Enrico Fermi postdoctoral scholar in the Mathematics and Computer Science Division. She received her Ph.D. and M.S. in computer science from the University of Tokyo in 2016 and 2012, respectively. Min’s research interests include irregular and dynamic communication models and runtime systems in high-performance computing.