Low power &amp; mobile hardware accelerators for deep convolutional neural networks

Scanlan, Anthony

doi:10.1016/j.vlsi.2018.11.010

Cited by 6 publications

(2 citation statements)

References 66 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By specifying the corresponding operation rules and transformation methods, and by expanding the basis of fuzzy mathematics, the fuzzy system model can be built based on quantitative method. In traditional mathematical theory, only 0 and 1 represents the logic relationship, the fuzzy mathematics can improve the traditional mathematics, quantifies the fuzzy relationship, and expresses the membership degree through number between 0 and 1, so as to deal with the fuzzy problem more effectively [13].…”

Section: Basic Theory Of Fuzzy Neural Networkmentioning

confidence: 99%

Research on Fault Diagnosis of Hydraulic System of Fast Erecting Device Based on Fuzzy Neural Network

Zheng

Xue

Zhang³

2022

TJFP

View full text Add to dashboard Cite

In order to improve the fault diagnosis effectiveness of hydraulic system in erecting devices, the fuzzy neural neural network is applied to carry out fault diagnosis of hydraulic system. Firstly, the main faults of hydraulic system of erecting mechanism are summarized. The main faults of hydraulic system of erecting devices concludes abnormal noise, high temperature of hydraulic oil of hydraulic system, leakage of hydraulic system, low operating speed of hydraulic system, and the characteristics of different faults are analyzed. Secondly, basic theory of fuzzy neural network is studied, and the framework of fuzzy neural network is designed. The inputting layer, fuzzy layer, fuzzy relation layer, relationship layer after fuzzy operation and outputting layer of fuzzy neural network are designed, and the corresponding mathematical models are confirmed. The analysis procedure of fuzzy neural network is established. Thirdly, simulation analysis is carried out for a hydraulic system in erecting device, the BP neural network reaches convergence after 600 times iterations, and the fuzzy neural network reaches convergence after 400 times iterations, fuzzy neural network can obtain higher accuracy than BP neural network, and running time of fuzzy neural network is less than that of BP neural network, therefore, simulation results show that the fuzzy neural network can effectively improve the fault diagnosis efficiency and precision. Therefore, the fuzzy neural network is reliable for fault diagnosis of hydraulic system in erecting devices, which has higher fault diagnosis effect, which can provide the theory basis for healthy detection of hydraulic system in erecting devices.

show abstract

Section: Basic Theory Of Fuzzy Neural Networkmentioning

confidence: 99%

Research on Fault Diagnosis of Hydraulic System of Fast Erecting Device Based on Fuzzy Neural Network

Zheng

Xue

Zhang³

2022

TJFP

View full text Add to dashboard Cite

show abstract

“…Dedicated hardware accelerators are extensively being advocated to be used in complex heterogeneous system-onchip to process large data more efficiently than pure software processing [1]. Moreover, hardware accelerates have a reduced power consumption, reduced latency and increased parallelism.…”

Section: Introductionmentioning

confidence: 99%

Machine Learning-Based Self-Compensating Approximate Computing

Masadeh¹,

Hasan²,

Tahar³

2020

Preprint

View full text Add to dashboard Cite

Dedicated hardware accelerators are suitable for parallel computational tasks. Moreover, they have the tendency to accept inexact results. These hardware accelerators are extensively used in image processing and computer vision applications, e.g., to process the dense 3-D maps required for self-driving cars. Such error-tolerant hardware accelerators can be designed approximately for reduced power consumption and/or processing time. However, since for some inputs the output errors may reach unacceptable levels, the main challenge is to enhance the accuracy of the results of approximate accelerators and keep the error magnitude within an allowed range. Towards this goal, in this paper, we propose a novel machine learning-based self-compensating approximate accelerators for energy efficient systems. The proposed error compensation module, which is integrated within the architecture of approximate hardware accelerators, efficiently reduces the accumulated error at its output. It utilizes lightweight supervised machine learning techniques, i.e., decision tree, to capture input dependency of the error. We consider image blending application in multiplication mode to demonstrate a practical application of self-compensating approximate computing. Simulation results show that the proposed design of self-compensating approximate accelerator can achieve about 9% accuracy enhancement, with negligible overhead in other performance measures, i.e., power, area, delay and energy.

show abstract

New paradigm of FPGA-based computational intelligence from surveying the implementation of DNN accelerators

Chang

et al. 2022

Des Autom Embed Syst

View full text Add to dashboard Cite

Low power & mobile hardware accelerators for deep convolutional neural networks

Cited by 6 publications

References 66 publications

Research on Fault Diagnosis of Hydraulic System of Fast Erecting Device Based on Fuzzy Neural Network

Research on Fault Diagnosis of Hydraulic System of Fast Erecting Device Based on Fuzzy Neural Network

Machine Learning-Based Self-Compensating Approximate Computing

New paradigm of FPGA-based computational intelligence from surveying the implementation of DNN accelerators

Contact Info

Product

Resources

About