Non-Volatile Memory and Associative Learning
Department
Electrical and Computer Engineering
Major
Computer Engineering & Computer Science
Research Advisor
Wunsch, Donald C.
Advisor's Department
Electrical and Computer Engineering
Funding Source
OURE
Abstract
Neuromorphic computing is a critical tool in modern problem solving, and non-volatile memory devices like memristors mitigate massive power consumption by transistor-based implementations. Memristors retain a set conductance level even with power off, enabling many practical applications. However, most research studies use idealized simulations, ignoring hardware implementations and non-ideal traits. This project investigates the use of commercially available hardware memristors and their non-ideal properties, to analyze associative learning applications. It demonstrates that non-ideal memristor components are not only feasible for use in machine learning applications, but can actually provide beneficial results when employed in associative memory algorithms.
Biography
Daniel Ellerbrock is a senior in Electrical and Computer Engineering graduating in May 2019. He has interests in non-volatile memory devices, computer architecture, and neuromorphic computing.When he graduates he will be working in the non-volatile memory systems group at Intel in California. On campus he has been involved in Kappa Alpha Order, Missouri S&T Jazz Band, and the Missouri S&T Chem E Car team. His undergraduate research is advised by Dr Donald Wunsch and is being conducted with fellow teammates Chris Rowan, Nicole Aldridge, and Yi Huang from Huazhong University of Science & Technology.
Research Category
Engineering
Presentation Type
Oral Presentation
Document Type
Presentation
Location
Ozark Room
Presentation Date
16 Apr 2019, 2:30 pm - 3:00 pm
Non-Volatile Memory and Associative Learning
Ozark Room
Neuromorphic computing is a critical tool in modern problem solving, and non-volatile memory devices like memristors mitigate massive power consumption by transistor-based implementations. Memristors retain a set conductance level even with power off, enabling many practical applications. However, most research studies use idealized simulations, ignoring hardware implementations and non-ideal traits. This project investigates the use of commercially available hardware memristors and their non-ideal properties, to analyze associative learning applications. It demonstrates that non-ideal memristor components are not only feasible for use in machine learning applications, but can actually provide beneficial results when employed in associative memory algorithms.
Comments
Joint project with Nicole Aldridge and Chris Rowan