MACHINE LEARNING PARAMETER TUNING

Connor Carroll (Junior)

LEARNING GOALS FOR RESEARCH IN PARAMETER TUNING

COMPLETED/IN PROGRESS

• Gain a comprehensive understanding of the structure and components of HPC machines and how they work together.
• Learn how to use shell scripting in the CentOS Linux environment to automate tasks and streamline workflows.
• Understand how a Linux kernel operates and gain knowledge of different Operating System structures.
• Learn about parallel computing and multithreading, including parallel programming models and parallel input-output in HPC.
• Gain practical experience in using Lustre and setting it up in a personal machine, including structuring a Lustre system and utilizing data sets.

​

LEARNING GOALS FOR RESEARCH IN PARAMETER TUNING

TO DO/IN PROGRESS

• Understand how hyperparameters in a Lustre system can affect system performance and analyze the effects of selected hyperparameters.
• Learn about optimization methods that utilize machine learning and DNN to automatically optimize hyperparameters while the system is running.
• Understand the efficiency of CAPES and test its performance in a personal environment.
• Review the efficiency of CAPES optimization on personal systems and analyze how its efficiency is affected by different system structures.
• Write/draw a poster to present findings at the Undergraduate Research Conference at UNC Charlotte.

​

SET UP OF SYSTEM/TESTING ENVIRONMENT

​

LUSTRE FILE MANAGEMENT SYSTEM

IMPORTANT CONSIDERATIONS AS I CONTINUE RESEARCH

CONSIDERATIONS