teaching

Duke University

ECE 662: Neuromorphic Computing & Machine Learning Acceleration

Teaching Assistant · Fall 2026 · Instructor: Prof. Hai (Helen) Li

A discussion-heavy, project-focused exploration of efficient AI systems, from model compression and accelerator architectures to emerging devices and spiking neural networks.

Modern AI models continue to grow in scale and computational demand, making efficient machine learning a central challenge in both research and industry. This graduate course examines how algorithms, architectures, circuits, and emerging devices can be co-designed to improve the efficiency of neural-network computation. Short lectures establish core concepts, while paper discussions, student presentations, design reviews, and a semester-long team project are where those concepts are tested and applied.

Topics

• Spiking neural networks (SNNs) and SNN accelerators
• Neural-network compression and low-precision quantization
• CPU/GPU acceleration and benchmarking
• FPGA-based acceleration and neural architecture search for FPGAs
• ASIC accelerators and extended ISAs
• Processing-in-memory (PIM), ReRAM, and other emerging-device accelerators
• Robustness of emerging device-based accelerators
• ML for edge and mobile computing
• ML for electronic design automation (EDA)

Guest lectures from Oak Ridge National Laboratory, HP Labs, Qualcomm, Google, and academia, covering SNN simulation, ReRAM-based accelerator design, and dynamic low-precision quantization.