Cross-layer CNN Approximations for Hardware Implementation

Karimi, Ali; Alouani, Ihsen; Ait El Cadi, Abdessamad; Ouarnoughi, Hamza; Niar, Smail

Document type :

Partie d'ouvrage

DOI :

10.1007/978-3-030-44534-8_12

Title :

Cross-layer CNN Approximations for Hardware Implementation

Author(s) :

Karimi, Ali [Auteur]
Laboratoire d'Automatique, de Mécanique et d'Informatique industrielles et Humaines - UMR 8201 [LAMIH]
Alouani, Ihsen [Auteur]
COMmunications NUMériques - IEMN [COMNUM - IEMN]
Institut d’Électronique, de Microélectronique et de Nanotechnologie - UMR 8520 [IEMN]
Ait El Cadi, Abdessamad [Auteur]
Laboratoire d'Automatique, de Mécanique et d'Informatique industrielles et Humaines - UMR 8201 [LAMIH]
Ouarnoughi, Hamza [Auteur]
Laboratoire d'Automatique, de Mécanique et d'Informatique industrielles et Humaines - UMR 8201 [LAMIH]
Niar, Smail [Auteur]
Laboratoire d'Automatique, de Mécanique et d'Informatique industrielles et Humaines - UMR 8201 [LAMIH]

Book title :

Applied Reconfigurable Computing. Architectures, Tools, and Applications

Publisher :

Springer International Publishing

Publication place :

Cham

Publication date :

2020-03-25

ISBN :

978-3-030-44533-1

English keyword(s) :

CNNs
FPGA
Approximate computing

HAL domain(s) :

Informatique [cs]

English abstract : [en]

Convolution Neural Networks (CNNs) are widely used for image classification and object detection applications. The deployment of these architectures in embedded applications is a great challenge. This challenge arises from ...
Show more >Convolution Neural Networks (CNNs) are widely used for image classification and object detection applications. The deployment of these architectures in embedded applications is a great challenge. This challenge arises from CNNs’ high computation complexity that is required to be implemented on platforms with limited hardware resources like FPGA. Since these applications are inherently error-resilient, approximate computing (AC) offers an interesting trade-off between resource utilization and accuracy. In this paper, we study the impact on CNN performances when several approximation techniques are applied simultaneously. We focus on two of the widely used approximation techniques, namely quantization and pruning. Our experimental results showed that for CNN networks of different parameter sizes and 3% loss in accuracy, we can obtain up to 27.9%–47.2% reduction in computation complexity in terms of FLOPs for CIFAR-10 and MNIST datasets.Show less >

Language :

Anglais

Audience :

Internationale

Popular science :

Non

Collections :

Institut d'Électronique, de Microélectronique et de Nanotechnologie (IEMN) - UMR 8520

Source :

Harvested from HAL

Cross-layer CNN Approximations for Hardware ... BibTeX CSV Excel RIS

Cross-layer CNN Approximations for Hardware ...

BibTeX

CSV

Excel

RIS