Shockers and HPC
This is a collection of just some of the HPC projects going on at 萝莉社.
Mathew Muether
Prof. Mathew Muether with physics masters students, Zak Elkarghli (MS graduate), Momin Khan (MS graduate), and Abdul Was it Yahaya is developing deep learning-based particle reconstruction tools for the international NOvA Neutrino Experiment based at Fermilab. Using BeoShock, the group has demonstrated that its techniques are significantly more accurate than traditional reconstruction tools. Prof. Muether was recently awarded a grant from the Department of Energy to continue this research. The high availability research grade GPU cluster in BeoShock is critical to this effort as it allows us flexibility to test and prototype before deploying tool at scale on the huge datasets provided by NOvA.
Katie Mitchell-Koch
Katie Mitchell-Koch: Beoshock has enabled successful completion of projects supported by grants that were already in existence, totaling $544,284. Moreover, Beoshock was featured as a vital resource in successful grant proposals that were funded during its time online, totaling $853,861, and it is featured in a pending grant proposal as well. Data from Beoshock has resulted in 8 publications since 2019, with more in preparation. I have generated calculations and data for projects with collaborators at University of Texas, Virginia Tech, University of Kansas, University of Minnesota, University of Basel, and Sardar Vallabhbhai National Institute of Technology, India (each collaborator/location is a distinct project). It has been the primary resource for research carried out by 6 graduate students and 2 postdoctoral researchers. Beoshock is used for students鈥� independent final projects using computational chemistry in CHEM 722, which is a core class in the graduate chemistry program.
Elizabeth Behrman and James Steck
Beoshock is important for the work of 萝莉社's quantum neural computing group, headed by Professor Elizabeth Behrman of math and physics and Professor James Steck of Aerospace Engineering. In this research we do quantum simulations of systems which themselves act as neural networks, "learning" to do the desired task, thus bypassing both quantum algorithm design and breakdown in terms of a series of gates. We have successfully trained a quantum system to estimate its own entanglement, and shown: that the additional learning necessary decreases with increasing size of system, that the calculation works in hardware as well as simulation, and that the entanglement estimation is resilient to both noise and decoherence. Currently we are doing learning of robust quantum repeaters, for long-distance quantum communication, and of hyperplanes bounding the unentangled subspaces. Beoshock is crucial when we simulate sytems over 6 qubits as the size of matrices in the calculations increases exponentially with system size. These are pushed onto the GPUs on GPU nodes, and the number of data points needed to be processed increases polynomially. These are done in parallel on the multiple cores on a node.
Nick Solomey
"The HPC system Beoshock is allowing my graduate students to explore potential new science that can be done with our new NASA Spacecraft development idea for a neutrino or dark matter detector in space, without it our progress of simulations would have been very limited"
Nick recommends these articles to learn more:
James Beck , Associate Professor, Biological Sciences
"I use High Performance Computing frequently to advance both my research and teaching. I run a plant diversity and evolution lab, and much of the data we collect these days is 'big' - data at thousands to tens of thousands of places in the genome. My students and I perform a variety of analyses on these data using the Beocat HPC at Kansas State University. I also use Beocat as part of my 'Computing for Biologists' course here in the Biology Department. Students in that course learn Unix, Python and R coding, and connect to and use Beocat remotely. High Performance Computing has greatly amplified what my students and I have been able to do.
Sergio Salinas Monroy: Community Data Labs
"Educators from KU, Kansas State University and 萝莉社are preparing new courses for the spring: Community Data Labs. These labs will bring together students from various disciplines to hone their data science skills by analyzing real-world information provided by partner organizations. Community partners are businesses, nonprofits and government organizations who will receive data-based solutions, generated by students, free of charge."
Want to have your projects included here? Email John Jones at john.jones@wichita.edu