IEEE CEDA Distinguished Lecture – HiKonv by Jinjun Xiong | 26 March 2026
Dear IEEE CEDA Members,
We are pleased to invite you to an IEEE CEDA Distinguished Lecture (DL) to be held on 26 March 2026, from 15:45 to 17:00, as part of our technical activities.
Distinguished Lecture Title
HiKonv: High Throughput Quantized Convolution on Conventional Multipliers with Novel Bit-wise Management and Computation
Speaker
Jinjun Xiong
University at Buffalo, USA
Abstract
The world of edge AI with constrained hardware resources has seen significant progress in applying quantization for deep Convolutional Neural Networks (CNNs) to reduce computation and storage costs using low-bitwidth data. However, there is still a lack of systematic studies on how existing full-bitwidth processing units—such as CPUs and DSPs—can be efficiently utilized to achieve significantly higher computation throughput for convolution across various quantized bitwidths.
In this distinguished lecture, we present HiKonv, a unified solution framework designed to maximize computation throughput on a given processing unit when executing low-bitwidth quantized CNNs. HiKonv is based on a novel bit-wise management strategy combined with highly parallel computation. The talk will introduce theoretical performance bounds for using full-bitwidth multipliers in parallelized low-bitwidth convolution and demonstrate new breakthroughs for high-performance CNN computation in this critical domain.
Venue
Thabraca Thalasso & Diving Hotel
We warmly encourage IEEE CEDA members, researchers, and practitioners interested in edge AI, hardware-efficient deep learning, and high-performance computing to attend this distinguished lecture.
We look forward to your participation.
Kind regards,
IEEE CEDA Tunisia Chapter Chair
Date and Time
Location
Hosts
Registration
-
Add Event to Calendar
- Tabarka, Jendouba
- Tunisia
- Starts 28 February 2026 11:00 PM UTC
- Ends 25 March 2026 11:00 PM UTC
- No Admission Charge