Soft errors in DNN accelerators: A comprehensive review

Ibrahim, Younis; Wang, Haibin; Liu, Junyang; Wei, Jinghe; Chen, Li; Rech, Paolo; Adam, Khalid; Guo, Gang

doi:10.1016/j.microrel.2020.113969

Cited by 61 publications

(23 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Interestingly, we found out the same results with our methodology as shown in next sections. In [27,28], the authors evaluate the reliability of one CNN executed on three different GPU architectures (Kepler, Maxwell, and Pascal).…”

Section: Related Workmentioning

confidence: 99%

Investigating data representation for efficient and reliable Convolutional Neural Networks

Ruospo

Sánchez

Traiola

et al. 2021

Microprocessors and Microsystems

View full text Add to dashboard Cite

Section: Related Workmentioning

confidence: 99%

Investigating data representation for efficient and reliable Convolutional Neural Networks

Ruospo

Sánchez

Traiola

et al. 2021

Microprocessors and Microsystems

View full text Add to dashboard Cite

“…On the other hand, there are some reports (for example references [ 137 , 138 , 139 ]) suggesting that some internal redundancy of computations embedded in AI models will suppress the negative effects of soft errors to some extent. Despite this, it is difficult to predict the levels of such error suppression with any degree of reliability.…”

Section: Resultsmentioning

confidence: 99%

Sources of Risk of AI Systems

Steimers

Schneider

2022

IJERPH

View full text Add to dashboard Cite

Artificial intelligence can be used to realise new types of protective devices and assistance systems, so their importance for occupational safety and health is continuously increasing. However, established risk mitigation measures in software development are only partially suitable for applications in AI systems, which only create new sources of risk. Risk management for systems that for systems using AI must therefore be adapted to the new problems. This work objects to contribute hereto by identifying relevant sources of risk for AI systems. For this purpose, the differences between AI systems, especially those based on modern machine learning methods, and classical software were analysed, and the current research fields of trustworthy AI were evaluated. On this basis, a taxonomy could be created that provides an overview of various AI-specific sources of risk. These new sources of risk should be taken into account in the overall risk assessment of a system based on AI technologies, examined for their criticality and managed accordingly at an early stage to prevent a later system failure.

show abstract

“…YOLOv3 was selected as the method for deep learning using SqueezeNet as a feature extractor. The rectified linear unit (ReLU) function used in fire modules, kept the original SqueezeNet activation settings [19], [29]. The fully connected (FC) layers will follow the leaky ReLU function.…”

Section: Methods 31 Proposed Approach In Deep Action Learningmentioning

confidence: 99%

Troop camouflage detection based on deep action learning

Muslikhin

Nasuha

Arifin

et al. 2022

IJ-AI

View full text Add to dashboard Cite

<p><span lang="EN-US">Detecting troop camouflage on the battlefield is crucial to beat or decide in critical situations to survive. This paper proposed a hybrid model based on deep action learning for camouflage recognition and detection. To involve deep action learning in this proposed system, deep learning based on you only look once (YOLOv3) with SquezeeNet and the fourth steps on action learning were engaged. Following the successful formulation of the learning cycle, an instrument examines the environment and performance in action learning with qualitative weightings; specific target detection experiments with view angle, target localization, and the firing point procedure were performed. For each deep action learning cycle, the complete process is divided into planning, acting, observing, and reflecting. If the results do not meet the minimal passing grade after the first cycle, the cycle will be repeated until the system succeeds in the firing point. Furthermore, this study found that deep action learning could enhance intelligence over earlier camouflage detection methods, while maintaining acceptable error rates. As a result, deep action learning could be used in armament systems if the environment is properly identified.</span></p>

show abstract

Soft errors in DNN accelerators: A comprehensive review

Cited by 61 publications

References 51 publications

Investigating data representation for efficient and reliable Convolutional Neural Networks

Investigating data representation for efficient and reliable Convolutional Neural Networks

Sources of Risk of AI Systems

Troop camouflage detection based on deep action learning

Contact Info

Product

Resources

About