TY - JOUR

T1 - TiQSA

T2 - Workload Minimization in Convolutional Neural Networks Using Tile Quantization and Symmetry Approximation

AU - Sabir, Dilshad

AU - Hanif, Muhammmad Abdullah

AU - Hassan, Ali

AU - Rehman, Saad

AU - Shafique, Muhammad

N1 - Funding Information:
This work was supported by the National University of Sciences and Technology, Islamabad.
Publisher Copyright:
© 2013 IEEE.

PY - 2021

Y1 - 2021

N2 - Convolutional Neural Networks (CNNs) in the Internet-of-Things (IoT)-based applications face stringent constraints, like limited memory capacity and energy resources due to many computations inconvolution layers. In order to reduce the computational workload in these layers, this paper proposes a hybrid convolution method in conjunction with a Particle of Swarm Convolution Layer Optimization (PSCLO) algorithm. The hybrid convolution is an approximation that exploits the inherent symmetry of filter termed as symmetry approximation and Winograd algorithm structure termed as tile quantization approximation. PSCLO optimizes the balance between workload reduction and accuracy degradation for each convolution layer by selecting fine-tuned thresholds to control each approximation's intensity. The proposed methods have been evaluated on ImageNet, MNIST, Fashion-MNIST, SVHN, and CIFAR-10 datasets. The proposed techniques achieved 5.28\text{x} multiplicative workload reduction without significant accuracy degradation (<0.1%) for ImageNet on ResNet-18, which is 1.08\text{x} less multiplicative workload as compared to state-of-the-art Winograd CNN pruning. For LeNet 3.87\text{x} and 3.93\text{x} was the multiplicative workload reduction for MNIST and Fashion-MNISTdatasets. The additive workload reduction was 2.5 {x} and 2.56 {x} for the respective datasets. There is no significant accuracy loss for MNIST and Fashion-MNIST dataset.

AB - Convolutional Neural Networks (CNNs) in the Internet-of-Things (IoT)-based applications face stringent constraints, like limited memory capacity and energy resources due to many computations inconvolution layers. In order to reduce the computational workload in these layers, this paper proposes a hybrid convolution method in conjunction with a Particle of Swarm Convolution Layer Optimization (PSCLO) algorithm. The hybrid convolution is an approximation that exploits the inherent symmetry of filter termed as symmetry approximation and Winograd algorithm structure termed as tile quantization approximation. PSCLO optimizes the balance between workload reduction and accuracy degradation for each convolution layer by selecting fine-tuned thresholds to control each approximation's intensity. The proposed methods have been evaluated on ImageNet, MNIST, Fashion-MNIST, SVHN, and CIFAR-10 datasets. The proposed techniques achieved 5.28\text{x} multiplicative workload reduction without significant accuracy degradation (<0.1%) for ImageNet on ResNet-18, which is 1.08\text{x} less multiplicative workload as compared to state-of-the-art Winograd CNN pruning. For LeNet 3.87\text{x} and 3.93\text{x} was the multiplicative workload reduction for MNIST and Fashion-MNISTdatasets. The additive workload reduction was 2.5 {x} and 2.56 {x} for the respective datasets. There is no significant accuracy loss for MNIST and Fashion-MNIST dataset.

KW - CNN

KW - Convolutional neural network

KW - DNN

KW - particle of swarm convolution layer optimization

KW - reduced workload

KW - symmetry approximation

KW - tile quantization approximation

KW - winograd transform

UR - http://www.scopus.com/inward/record.url?scp=85103755552&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85103755552&partnerID=8YFLogxK

U2 - 10.1109/ACCESS.2021.3069906

DO - 10.1109/ACCESS.2021.3069906

M3 - Article

AN - SCOPUS:85103755552

SN - 2169-3536

VL - 9

SP - 53647

EP - 53668

JO - IEEE Access

JF - IEEE Access

M1 - 9389774

ER -