DNN Accelerator

DNN accelerators are specialized hardware designed to efficiently execute deep neural network (DNN) computations, primarily aiming to improve speed, reduce energy consumption, and minimize latency. Current research focuses on optimizing various aspects of these accelerators, including novel memory hierarchies, efficient in-memory computing (IMC) using stochastic processing, and adaptive hardware/software co-optimization techniques, often applied to models like ResNet and Vision Transformers. These advancements are crucial for deploying DNNs on resource-constrained edge devices and in safety-critical applications, impacting both the efficiency of AI systems and their reliability in real-world deployments.

Papers

July 16, 2023

MaGNAS: A Mapping-Aware Graph Neural Architecture Search Framework for Heterogeneous MPSoC Deployment
Mohanad Odema, Halima Bouzidi, Hamza Ouarnoughi, Smail Niar, Mohammad Abdullah Al Faruque
Graph Neural Network DNN Accelerator DNN Workload Graph Neural Architecture Search Heterogeneous System on Chip

June 16, 2023

Enhancing Fault Resilience of QNNs by Selective Neuron Splitting
Mohammad Hasan Ahmadilivani, Mahdi Taheri, Jaan Raik, Masoud Daneshtalab, Maksim Jenihhin
Neural Network DNN Accelerator Quantized Neural Network Critical Neuron Fault Resilience

June 14, 2023

MetaML: Automating Customizable Cross-Stage Design-Flow for Deep Learning Acceleration
Zhiqiang Que, Shuo Liu, Markus Rognlien, Ce Guo, Jose G. F. Coutinho, Wayne Luk
Deep Neural Network Hardware Accelerator DNN Accelerator Deep Learning Acceleration

June 12, 2023

On the Viability of using LLMs for SW/HW Co-Design: An Example in Designing CiM DNN Accelerators
Zheyu Yan, Yifan Qin, Xiaobo Sharon Hu, Yiyu Shi
Large Language Model Deep Neural Network DNN Model Co Design DNN Accelerator Hardware Design Myocardial Viability

May 31, 2023

Special Session: Approximation and Fault Resiliency of DNN Accelerators
Mohammad Hasan Ahmadilivani, Mario Barbareschi, Salvatore Barone, Alberto Bosio, Masoud Daneshtalab, Salvatore Della Torca, Gabriele Gavarini, Maksim Jenihhin, Jaan Raik, Annachiara Ruospo, Ernesto Sanchez, Mahdi Taheri
Deep Learning Deep Neural Network Average Approximation DNN Accelerator Special Session Fault Resilience

May 30, 2023

NicePIM: Design Space Exploration for Processing-In-Memory DNN Accelerators with 3D-Stacked-DRAM
Junpeng Wang, Mengke Ge, Bo Ding, Qi Xu, Song Chen, Yi Kang
DNN Accelerator Design Space Exploration Processing in Memory

May 25, 2023

PQA: Exploring the Potential of Product Quantization in DNN Hardware Acceleration
Ahmed F. AbouElhamayed, Angela Cui, Javier Fernandez-Marques, Nicholas D. Lane, Mohamed S. Abdelfattah
Full Potential DNN Accelerator Quantized Neural Network Product Quantization

May 23, 2023

Negative Feedback Training: A Novel Concept to Improve Robustness of NVCIM DNN Accelerators
Yifan Qin, Zheyu Yan, Wujie Wen, Xiaobo Sharon Hu, Yiyu Shi
Native Robustness DNN Accelerator DNN Inference Non Volatile Memory DNN Robustness Novel Concept Device Variation Negative Training

May 22, 2023

May 21, 2023

FAQ: Mitigating the Impact of Faults in the Weight Memory of DNN Accelerators through Fault-Aware Quantization
Muhammad Abdullah Hanif, Muhammad Shafique
Global Impact DNN Accelerator Weight Freezing Post Fault Trajectory Fault Aware

May 10, 2023

MoCA: Memory-Centric, Adaptive Execution for Multi-Tenant Deep Neural Networks
Seah Kim, Hasan Genc, Vadim Vadimovich Nikiforov, Krste Asanović, Borivoje Nikolić, Yakun Sophia Shao
DNN Accelerator Multiple Deep Neural Network Efficient Execution Latency Critical Multi Tenant Resource Contention

April 20, 2023

March 14, 2023

DeepAxe: A Framework for Exploration of Approximation and Reliability Trade-offs in DNN Accelerators
Mahdi Taheri, Mohammad Riazati, Mohammad Hasan Ahmadilivani, Maksim Jenihhin, Masoud Daneshtalab, Jaan Raik, Mikael Sjodin, Bjorn Lisper
New Framework Average Approximation Pareto Optimal DNN Accelerator Trade Offs High Level Synthesis

January 12, 2023

Improving Reliability of Spiking Neural Networks through Fault Aware Threshold Voltage Optimization
Ayesha Siddique, Khaza Anuarul Hoque
Spiking Neural Network Neuromorphic Hardware Neuromorphic Datasets DNN Accelerator

December 5, 2022

Thales: Formulating and Estimating Architectural Vulnerability Factors for DNN Accelerators
Abhishek Tyagi, Yiming Gan, Shaoshan Liu, Bo Yu, Paul Whatmough, Yuhao Zhu
DNN Accelerator New Formulation DNN Accuracy Resilience Measure Transient Fault

October 10, 2022

Energy-Efficient Deployment of Machine Learning Workloads on Neuromorphic Hardware
Peyton Chandarana, Mohammadreza Mohammadi, James Seekings, Ramtin Zand
Neuromorphic Hardware Hardware Accelerator DNN Accelerator Machine Learning Workload Efficient Deployment Neuromorphic Research Chip Loihi

August 30, 2022

ANT: Exploiting Adaptive Numerical Data Type for Low-bit Deep Neural Network Quantization
Cong Guo, Chen Zhang, Jingwen Leng, Zihan Liu, Fan Yang, Yunxin Liu, Minyi Guo, Yuhao Zhu
DNN Accelerator Low Bit Quantization Quantized Neural Network Dynamic Quantization Ant Colony Data Adaptation

July 25, 2022

Energy-efficient DNN Inference on Approximate Accelerators Through Formal Property Exploration
Ourania Spantidi, Georgios Zervakis, Iraklis Anagnostopoulos, Jörg Henkel
Energy Efficient DNN Accelerator DNN Accuracy Accuracy Trade Weight Re Mapping Approximate DNN Accelerator

DNN Accelerator

Papers

MaGNAS: A Mapping-Aware Graph Neural Architecture Search Framework for Heterogeneous MPSoC Deployment

Enhancing Fault Resilience of QNNs by Selective Neuron Splitting

MetaML: Automating Customizable Cross-Stage Design-Flow for Deep Learning Acceleration

On the Viability of using LLMs for SW/HW Co-Design: An Example in Designing CiM DNN Accelerators

Special Session: Approximation and Fault Resiliency of DNN Accelerators

NicePIM: Design Space Exploration for Processing-In-Memory DNN Accelerators with 3D-Stacked-DRAM

PQA: Exploring the Potential of Product Quantization in DNN Hardware Acceleration

Negative Feedback Training: A Novel Concept to Improve Robustness of NVCIM DNN Accelerators

ADA-GP: Accelerating DNN Training By Adaptive Gradient Prediction

HighLight: Efficient and Flexible DNN Acceleration with Hierarchical Structured Sparsity

FAQ: Mitigating the Impact of Faults in the Weight Memory of DNN Accelerators through Fault-Aware Quantization

MoCA: Memory-Centric, Adaptive Execution for Multi-Tenant Deep Neural Networks

SALSA: Simulated Annealing based Loop-Ordering Scheduler for DNN Accelerators

eFAT: Improving the Effectiveness of Fault-Aware Training for Mitigating Permanent Faults in DNN Hardware Accelerators

DeepAxe: A Framework for Exploration of Approximation and Reliability Trade-offs in DNN Accelerators

Improving Reliability of Spiking Neural Networks through Fault Aware Threshold Voltage Optimization

Thales: Formulating and Estimating Architectural Vulnerability Factors for DNN Accelerators

Energy-Efficient Deployment of Machine Learning Workloads on Neuromorphic Hardware

ANT: Exploiting Adaptive Numerical Data Type for Low-bit Deep Neural Network Quantization

Energy-efficient DNN Inference on Approximate Accelerators Through Formal Property Exploration