CNN Accelerator

CNN accelerators are specialized hardware designed to significantly speed up and reduce the energy consumption of convolutional neural network (CNN) inference, crucial for deploying AI on resource-constrained devices. Current research focuses on optimizing various aspects, including efficient data movement through novel memory mapping strategies and dataflow architectures, exploiting sparsity in network weights and activations via pruning and quantization techniques, and developing hardware-aware training methods. These advancements are vital for enabling real-time AI applications in areas like embedded systems and IoT, where low latency and power consumption are paramount.

Papers

September 8, 2024

BBS: Bi-directional Bit-level Sparsity for Deep Learning Acceleration
Yuzong Chen, Jian Meng, Jae-sun Seo, Mohamed S. Abdelfattah
Sparsity Increase CNN Accelerator Deep Learning Acceleration Sparsity Level

August 26, 2024

HAPM -- Hardware Aware Pruning Method for CNN hardware accelerators in resource constrained devices
Federico Nicolas Peccia, Luciano Ferreyro, Alejandro Furfaro
Convolutional Neural Network Hardware Accelerator New Resource Smart Device CNN Accelerator

August 5, 2024

PENDRAM: Enabling High-Performance and Energy-Efficient Processing of Deep Neural Networks through a Generalized DRAM Data Mapping Policy
Rachmad Vidya Wicaksana Putra, Muhammad Abdullah Hanif, Muhammad Shafique
Deep Neural Network High Performance CNN Accelerator Random Access Memory

July 24, 2024

SAfEPaTh: A System-Level Approach for Efficient Power and Thermal Estimation of Convolutional Neural Network Accelerator
Yukai Chen, Simei Yang, Debjyoti Bhattacharjee, Francky Catthoor, Arindam Mallik
CNN Accelerator Accelerator Architecture Power Optimization Analog Compute in Memory Fast Thermal

May 7, 2024

NOVA: NoC-based Vector Unit for Mapping Attention Layers on a CNN Accelerator
Mohit Upadhyay, Rohan Juneja, Weng-Fai Wong, Li-Shiuan Peh
Attention Mechanism Attention Layer Attention Map X Vector CNN Accelerator View Augmentation Attention Based Neural Network Neural Accelerator

March 14, 2024

FlexNN: A Dataflow-aware Flexible Deep Learning Accelerator for Energy-Efficient Edge Devices
Arnab Raha, Deepak A. Mathaikutty, Soumendu K. Ghosh, Shamik Kundu
Edge Device DNN Accelerator CNN Accelerator AlexNet Architecture Dataflow Based

March 8, 2024

SF-MMCN: Low-Power Sever Flow Multi-Mode Diffusion Model Accelerator
Huan-Ke Hsu, I-Chyn Wey, T. Hui Teo
Convolutional Neural Network Model CNN Accelerator

March 5, 2024

Mem-elements based Neuromorphic Hardware for Neural Network Application
Ankur Singh
Neuromorphic Hardware CNN Accelerator Complementary Metal Oxide Semiconductor Neural Network Application

December 21, 2023

Cross-Layer Optimization for Fault-Tolerant Deep Learning
Qing Zhang, Cheng Liu, Bo Liu, Haitong Huang, Ying Wang, Huawei Li, Xiaowei Li
Neural Network CNN Accelerator Cross Layer Deep Learning Accelerator Reliable Deep Learning

November 20, 2023

Improvements in Interlayer Pipelining of CNN Accelerators Using Genetic Algorithms
Mark Horeni, Siddharth Joshi
Neural Network Convolutional Neural Network Genetic Algorithm Large Relevance Improvement Hardware Acceleration L MobileNet CNN Accelerator

November 1, 2023

Revealing CNN Architectures via Side-Channel Analysis in Dataflow-based Inference Accelerators
Hansika Weerasena, Prabhat Mishra
CNN Architecture Side Channel CNN Accelerator Dataflow Based

July 25, 2023

Mitigating Memory Wall Effects in CNN Engines with On-the-Fly Weights Generation
Stylianos I. Venieris, Javier Fernandez-Marques, Nicholas D. Lane
Deep Learning CNN Accelerator Weight Generation

July 12, 2023

Flexible and Fully Quantized Ultra-Lightweight TinyissimoYOLO for Ultra-Low-Power Edge Systems
Julian Moosmann, Hanna Mueller, Nicky Zimmerman, Georg Rutishauser, Luca Benini, Michele Magno
Edge Device Hardware Accelerator Extension Study Detection Performance CNN Accelerator Lightweight Object Detection

June 22, 2023

To Spike or Not to Spike? A Quantitative Comparison of SNN and CNN FPGA Implementations
Patrick Plagwitz, Frank Hannig, Jürgen Teich, Oliver Keszocze
Energy Efficient Spike Protein Empirical Comparison CNN Accelerator SNN Architecture

May 22, 2023

TinyissimoYOLO: A Quantized, Low-Memory Footprint, TinyML Object Detection Network for Low Power Microcontrollers
Julian Moosmann, Marco Giordano, Christian Vogt, Michele Magno
TinyML Model CNN Accelerator Lightweight Object Detection Detection Class

December 7, 2022

CODEBench: A Neural Architecture and Hardware Accelerator Co-Design Framework
Shikhar Tuli, Chia-Hao Li, Ritvik Sharma, Niraj K. Jha
Neural Architecture Co Design CNN Accelerator Accelerator Architecture

December 6, 2022

Vision Transformer Computation and Resilience for Dynamic Inference
Kavya Sreedhar, Jason Clemons, Rangharajan Venkatesan, Stephen W. Keckler, Mark Horowitz
Vision Transformer Deep Learning Model Transformer Architecture Tiny Refinement Elicit Resilience Transformer Inference CNN Accelerator Dynamic Inference

November 10, 2022

PhotoFourier: A Photonic Joint Transform Correlator-Based Neural Network Accelerator
Shurui Li, Hangbo Yang, Chee Wei Wong, Volker J. Sorger, Puneet Gupta
CNN Accelerator Neural Network Accelerator Photonic Computing Joint Image Neuromorphic Photonic Spatial Accelerator

September 22, 2022

Optimization of FPGA-based CNN Accelerators Using Metaheuristics
Sadiq M. Sait, Aiman El-Maleh, Mohammad Altakrouri, Ahmad Shawahna
Optimization Purpose Field Programmable Gate Array Metaheuristic Algorithm CNN Accelerator Multi Accelerator

June 30, 2022

Sparse Periodic Systolic Dataflow for Lowering Latency and Power Dissipation of Convolutional Neural Network Accelerators
Jung Hwan Heo, Arash Fayyazi, Amirhossein Esmaili, Massoud Pedram
CNN Accelerator Sparsity Pattern Multi Accelerator Deep Learning Compiler Efficient Compilation Systolic Array Systolic Complex Power Dissipation

CNN Accelerator

Papers

BBS: Bi-directional Bit-level Sparsity for Deep Learning Acceleration

HAPM -- Hardware Aware Pruning Method for CNN hardware accelerators in resource constrained devices

PENDRAM: Enabling High-Performance and Energy-Efficient Processing of Deep Neural Networks through a Generalized DRAM Data Mapping Policy

SAfEPaTh: A System-Level Approach for Efficient Power and Thermal Estimation of Convolutional Neural Network Accelerator

NOVA: NoC-based Vector Unit for Mapping Attention Layers on a CNN Accelerator

FlexNN: A Dataflow-aware Flexible Deep Learning Accelerator for Energy-Efficient Edge Devices

SF-MMCN: Low-Power Sever Flow Multi-Mode Diffusion Model Accelerator

Mem-elements based Neuromorphic Hardware for Neural Network Application

Cross-Layer Optimization for Fault-Tolerant Deep Learning

Improvements in Interlayer Pipelining of CNN Accelerators Using Genetic Algorithms

Revealing CNN Architectures via Side-Channel Analysis in Dataflow-based Inference Accelerators

Mitigating Memory Wall Effects in CNN Engines with On-the-Fly Weights Generation

Flexible and Fully Quantized Ultra-Lightweight TinyissimoYOLO for Ultra-Low-Power Edge Systems

To Spike or Not to Spike? A Quantitative Comparison of SNN and CNN FPGA Implementations

TinyissimoYOLO: A Quantized, Low-Memory Footprint, TinyML Object Detection Network for Low Power Microcontrollers

CODEBench: A Neural Architecture and Hardware Accelerator Co-Design Framework

Vision Transformer Computation and Resilience for Dynamic Inference

PhotoFourier: A Photonic Joint Transform Correlator-Based Neural Network Accelerator

Optimization of FPGA-based CNN Accelerators Using Metaheuristics

Sparse Periodic Systolic Dataflow for Lowering Latency and Power Dissipation of Convolutional Neural Network Accelerators