A Scalable System-on-Chip Acceleration for Deep Neural Networks

Faisal Shehzad,Saud S Alotaibi,Muhammad Rashid,Mohammed H Sinky,Muhammad Yousuf Irfan Zia

doi:10.1109/access.2021.3094675

Abstract

The size of neural networks in deep learning techniques is increasing and varies significantly according to the requirements of real-life applications. The increasing network size and scalability requirements pose significant challenges for a high performance implementation of deep neural networks (DNN). Conventional implementations, such as graphical processing units and application specific integrated circuits, are either less efficient or less flexible. Consequently, this article presents a system-on-chip (SoC) solution for the acceleration of DNN, where an ARM processor controls the overall execution and off-loads computational intensive operations to a hardware accelerator. The system implementation is performed on a SoC development board. Experimental results show that the proposed system achieves a speed-up of 22.3, with a network architecture size of 64×64, in comparison with the native implementation on a dual core cortex ARM-A9 processor. In order to generalize the performance of complete system, a mathematical formula is presented which allows to compute the total execution time for any architecture size. The validation is performed by taking Epileptic Seizure Recognition as the target case study. Finally, the results of the proposed solution are compared with various state-of-the-art solutions in terms of execution time, scalability, and clock frequency.

Highlights

Deep learning algorithms are getting increasingly popular for image classification [1], object detection [2] and data prediction [3] in numerous real-world applications
EVALUATION RESULTS This section evaluates the performance of proposed Hardware Accelerator for Deep Learning (HADL) system for various architecture sizes in terms of execution time, accuracy, clock frequency, power consumption and the required hardware resources
The proposed co-design approach starts with the construction of a deep neural networks (DNN) for the target case study

Summary

Introduction

Deep learning algorithms are getting increasingly popular for image classification [1], object detection [2] and data prediction [3] in numerous real-world applications. It has dramatically increased the development speed of machine learning (ML) and artificial intelligence (AI) [4]. The basic concept behind all the deep learning applications is to use a multilayer neural network model for the extraction of high-level features These high-level features combine various low-level abstractions to find some distributed data features [7]. The EEG is a clinical process that monitors the activity of the human brain while performing some cognitive task

Objectives

Methods

Results

Conclusion