Presenter(s)

Yassine Jaoudi

Download

Download Project (191 KB)

Description

Training deep learning models is computationally expensive due to the need for a tremendous volume of data and complex math. Graphical Processing Units (GPUs) are typically used and require about 200W of power at least, thus making them unusable in portable applications. Neuromorphic computing approaches based on memristor devices can drastically reduce this power and allow low power devices (edge computing and IoT devices) to learn and thus become much smarter. This work presents collected characteristics data of real memristor devices and modeling for memristor-based circuit and system design. Memristors – a relatively recent class on nanoscale devices that can be programmed and can retain their data even when the power is turned off. Memristor based online circuits is a popular research topic currently, but these are generally based on ideal devices behaviors. Therefore, the acquired device properties are used to update the memristor model used in previous circuit simulations and examine its impact on Artificial Intelligence learning circuits.

Publication Date

4-22-2020

Project Designation

Graduate Research

Primary Advisor

Tarek M. Taha

Primary Advisor's Department

Electrical and Computer Engineering

Keywords

Stander Symposium project, School of Engineering

Electrical Characterization of Tantalum Oxide Based Memristor

Share

COinS