MRI: Acquisition of Adaptive Cluster for Performance and Forensics Analysis of Distributed Machine Learning

点击翻译按钮获取中文摘要

This project, acquiring a computational cluster, aims to provide opportunities for research in machine-learning, algorithm development, and protection of information in multiple environments. The capacity to evaluate and analyze performance and residue data generation in data mining, machine learning computations, should allow better control and less risk of breaches in cybersecurity. The availability of these enhancements would also permit use of these systems for applied, interdisciplinary research using large-scale data and cross-correlation analyses for predictive modeling. The investigators measure performance systematically to support forensic analysis of data residues, in order to detect possible security risks in the use of such platforms. The procurement of the instrumentation yields a significant expansion in data mining, security, and forensics research. Core research foci in cyber security, digital forensics, and data mining research enables a work plan based on defined problems in distributed computing environments related to performance, algorithms, and data security. The gained instrument and expertise provide the institution with the ability to support national level customers such as U.S. Army Aviation & Missile Research, Development and Engineering Center, National Institute of Health (NIH), and other government agencies that depend on effective and secure distributed learning to analyze and process sensitive data. A key issue with solving both the forensics challenges and the performance analysis is having access to an instrument such that investigators can- Tune, adjust, and redeploy environments, - Take nodes offline to be examined forensically, - Ensure a consistent baseline exists against which other environments are compared, and - Run meaningful experiments within their discipline while enabling the collection of valuable performance and forensics data (permitting non-data mining and utilizing digital forensics).The proposed adaptive cluster instrument enables these four goals.Broader Impacts:The computational research capabilities provide essential resources for undergraduate and graduate students' research as well as for training and exposure activities involving K-12. Students will be afforded hands-on opportunities in research and classroom activities directly utilizing the cluster. The skills gained should lead directly to internships and permanent employment opportunities for doctoral students in the six colleges and/or institutes. The university services a high percentage of rural and financially depressed areas; has a 34% enrollment of non-white students, with 60% female enrollment. The instrumentation offers exposure opportunities for cohort activities to Scholarships for Service Program.