Collaborative Research: CyberTraining: Pilot: Research Workforce Development for Deep Learning Systems in Advanced GPU Cyberinfrastructure

Overview With the recent advancements in artificial intelligence, deep learning (DL) systems and applications have become a driving force in multiple transdisciplinary domains. Meanwhile, this evolution has been supported by the rapid improvements of advanced GPU cyberinfrastructure (CI). Thus, DL system and application researchers need to constantly acquire interdisciplinary knowledge on advanced GPU CI. However, existing training materials are not only separately designed for a single aspect, either GPU CI or DL systems and applications, but also mainly focused on traditional techniques lagging behind quickly developed GPU CI and DL systems. Moreover, developing a comprehensive multidisciplinary curriculum for interdisciplinary research on GPU-based DL systems is a challenging task due to current disciplinary training, knowledge limitation of a single instructor, and nonextensive training on advanced GPU CI, especially in rural areas. To fill in this blank and overcome these difficulties, this project will develop a novel set of interactive training materials by involving six faculty members from five academic disciplines including computer science, computer engineering, data science, geospatial information science, and aerospace engineering. Particularly, the training material will include hands-on lecture modules, invited relevant research talks, and an openended collaborative project in an intensive interdisciplinary workshop, integrating interconnected cuttingedge techniques in advanced GPU CI for DL systems. The goal of this two-week online workshop is to foster future CI users and contributors, who can use, develop, and improve advanced GPU CI for DL systems in their research. We anticipate that the workshop will enable participants, including seniors, graduate students, and researchers, to refine their multidisciplinary skillsets, extend their academic research portfolios, develop their remote collaboration capacities, and significantly strengthen their career competitiveness. Keywords: CI contributors and users; seniors, graduate students, and researchers; advanced CI, computer and network systems, geosciences; advanced GPU CI, DL systems and applications Intellectual Merit The intellectual merit of this project is mainly three-fold. (i) Comprehensive understanding on advanced GPU CI for DL: lecture modules will provide trainees with knowledge and skills on the full stack of DL systems in advanced GPU CI, including state-of-the-art GPU architectures, advanced CUDA toolkits for DL, GPU-based DL systems, and DL application benchmarks in geoscience research to foster an educational ecosystem for interdisciplinary research. (ii) Remote interdisciplinary collaboration: an open-ended interdisciplinary project will provide participants with an opportunity to remotely work together with team members from different disciplines to leverage their complementary background. (iii) Efficient online training of interactive hands-on exercises: an interactive training prototype system will be developed to automatically detect the completion of trainees’ hands-on exercises on PIs’ high-end GPU servers, so that teaching speed can be dynamically adjusted according to trainees’ learning progress. Broader Impacts (i) Teaching materials will be incorporated into nine existing courses in five departments across five academic disciplines at SIUC and Missouri S&T, and as the cornerstone for developing a new one used to develop one. (ii) They will also be made publicly available for adoption by other institutions, with guidelines provided. (iii) The designed project will enable future scientific and engineering research workforce to effectively use and develop advanced GPU CI for DL systems for their scientific innovation and discovery. (iv) PIs will invite speakers from GPU CI job sectors (such as research laboratories of U.S. Department of Energy and multinational corporations) for research talks to create opportunities in close interaction between participants and job providers, bridging the gaps between these two communities. (v) PIs will broadly disseminate the project outcome through various conferences and outreach activities for educational and research communities. (vi) Minority students and rural researchers will be recruited into our workshops as a high priority to increase the diversity of research workforce and make up the CI workforce gap between rural and urban areas, respectively. This project will also form a coordination network consisting of research universities and minority-serving institutions.