ELEG/CISC x87-022: VIP: Neuromorphic Computing

Fall 2020, University of Delaware

 

TL;DR: This vertically integrated project (VIP) course focuses on building artificial intelligence devices with brain-like energy-efficiency.

 

For an overview of the VIP program at UD see refer to the site: https://vip.udel.edu/

 

VIP Advisor               : Dr. Vishal Saxena (Website)

Email                          : vsaxena@udel.edu   

Class Schedule           : We 5:40PM–6:30PM

Class dates                 : Sep 1– Dec 10, 2020

Office Hours              : Tue & Thu 2-3 PM

Slack Channel           : TBD 

 

Textbook – None. Datasheets, papers and documents will be provided as needed.

 

Course content (Syllabus) –

https://vip.udel.edu/resource/VIP%20Syllabus

 

Goals: Develop new generation of computing hardware and algorithms that can realized Artificial Intelligence (AI) on portable edge devices. Emphasis will be on using neural-inspired, or Neuromorphic Architectures, that can allow brain-like energy efficiency. The devices would be able to process vision, speech, text, and sensor signals, and make decisions based on learned behavior.

Key Elements: Neuromorphic Computing, Artificial Intelligence at the Edge; Deep learning, Artificial Neural Networks, Development of CMOS circuits using emerging devices, In-Memory Computing, Implementation of learning algorithms on available Edge-AI/neuromorphic hardware platforms and FPGAs.

 

Research Issues: Development of neuromorphic computers involves challenges at the electronic devices, circuits, and algorithm level. Not only each of these components have their own research challenges, and integrative approach is followed where they are synergistically developed. This research encompasses a range of projects including:

1.      Spiking neural network (SNN) algorithm development on CPU/GPU, available neuromorphic hardware platforms, and FPGAs

2.      Development of novel circuits using emerging memory devices such as RRAM, PCRAM and spintronics

3.      Development of custom Neuromorphic Systems-on-a-Chip (SoCs) using CMOS technology integrated with memory devices

4.      Sensor development and interfacing with neuromorphic platforms.

 

Major, Preparation and Interests: Computer Engineering – Integrated Circuits, Computer Architecture, Embedded Systems; Electrical Engineering – Machine Learning, Signal Processing; Computer Science – Neural Networks, Algorithms, Programming/Software; Biomedical Engineering – Sensor readout and signal processing.

 

Research Areas: Computer Engineering, Integrated Circuits and Systems, Machine Learning, Computational Neuroscience

 

Prerequisites – Instructor permission.  

 

Workload (Grading) 

100% Project report

 

 

Return