Algorithm and Hardware Co-design for Reconfigurable CNN Accelerator

11/24/2021
by   Hongxiang Fan, et al.
9

Recent advances in algorithm-hardware co-design for deep neural networks (DNNs) have demonstrated their potential in automatically designing neural architectures and hardware designs. Nevertheless, it is still a challenging optimization problem due to the expensive training cost and the time-consuming hardware implementation, which makes the exploration on the vast design space of neural architecture and hardware design intractable. In this paper, we demonstrate that our proposed approach is capable of locating designs on the Pareto frontier. This capability is enabled by a novel three-phase co-design framework, with the following new features: (a) decoupling DNN training from the design space exploration of hardware architecture and neural architecture, (b) providing a hardware-friendly neural architecture space by considering hardware characteristics in constructing the search cells, (c) adopting Gaussian process to predict accuracy, latency and power consumption to avoid time-consuming synthesis and place-and-route processes. In comparison with the manually-designed ResNet101, InceptionV2 and MobileNetV2, we can achieve up to 5 with other state-of-the-art co-design frameworks, our found network and hardware configuration can achieve 2 latency and 8.5x higher energy efficiency.

READ FULL TEXT

page 1

page 3

research
04/26/2021

HAO: Hardware-aware neural Architecture Optimization for Efficient Inference

Automatic algorithm-hardware co-design for DNN has shown great success i...
research
08/01/2021

FLASH: Fast Neural Architecture Search with Hardware Optimization

Neural architecture search (NAS) is a promising technique to design effi...
research
06/23/2021

NAX: Co-Designing Neural Network and Hardware Architecture for Memristive Xbar based Computing Systems

In-Memory Computing (IMC) hardware using Memristive Crossbar Arrays (MCA...
research
09/14/2020

AutoML for Multilayer Perceptron and FPGA Co-design

State-of-the-art Neural Network Architectures (NNAs) are challenging to ...
research
09/14/2020

DANCE: Differentiable Accelerator/Network Co-Exploration

To cope with the ever-increasing computational demand of the DNN executi...
research
11/21/2018

HAQ: Hardware-Aware Automated Quantization

Model quantization is a widely used technique to compress and accelerate...
research
01/23/2023

Enabling Hard Constraints in Differentiable Neural Network and Accelerator Co-Exploration

Co-exploration of an optimal neural architecture and its hardware accele...

Please sign up or login with your details

Forgot password? Click here to reset