scholarly journals Low-Voltage Energy Efficient Neural Inference by Leveraging Fault Detection Techniques

2021 ◽  
Author(s):  
Mehdi Safarpour ◽  
Tommy Z. Deng ◽  
John Massingham ◽  
Lei Xun ◽  
Mohammad Sabokrou ◽  
...  
Keyword(s):  
Fault Detection ◽  
Error Detection ◽  
Energy Savings ◽  
Low Voltage ◽  
Building Blocks ◽  
Feedback Mechanism ◽  
Total System ◽  
Data Set ◽  
Detection Techniques ◽  
Energy Efficiency Improvement

Operating at reduced voltages offers substantial energy efficiency improvement but at the expense of increasing the probability of computational errors due to hardware faults. In this context, we targeted Deep Neural Networks (DNN) as emerging energy hungry building blocks in embedded applications. Without an error feedback mechanism, blind voltage down-scaling will result in degraded accuracy or total system failure. To enable safe voltage down-scaling, in this paper two solutions based on Self-Supervised Learning (SSL) and Algorithm Based Fault Tolerance (ABFT) were developed. A DNN model trained on MNIST data-set was deployed on a Field Programmable Gate Array (FPGA) that operated at reduced voltages and employed the proposed schemes. The SSL approach provides extremely low-overhead (≈0.2%) fault detection at the cost of lower error coverage and extra training, while ABFT incurs less than 8%overheads at run-time with close to 100% error detection rate. By using the solutions, substantial energy savings, i.e., up to 40.3%,without compromising the accuracy of the model was achieved

scholarly journals Low-Voltage Energy Efficient Neural Inference by Leveraging Fault Detection Techniques

2021 ◽  
Author(s):  
Mehdi Safarpour ◽  
Tommy Z. Deng ◽  
John Massingham ◽  
Lei Xun ◽  
Mohammad Sabokrou ◽  
...  
Keyword(s):  
Fault Detection ◽  
Error Detection ◽  
Energy Savings ◽  
Low Voltage ◽  
Building Blocks ◽  
Feedback Mechanism ◽  
Total System ◽  
Data Set ◽  
Detection Techniques ◽  
Energy Efficiency Improvement

This paper presents simple techniques to significantly reduced energy consumption of DNNs: Operating at reduced voltages offers substantial energy efficiency improvement but at the expense of increasing the probability of computational errors due to hardware faults. In this context, we targeted Deep Neural Networks (DNN) as emerging energy hungry building blocks in embedded applications. Without an error feedback mechanism, blind voltage down-scaling will result in degraded accuracy or total system failure. To enable safe voltage down-scaling, in this paper two solutions based on Self-Supervised Learning (SSL) and Algorithm Based Fault Tolerance (ABFT) were developed. A DNN model trained on MNIST data-set was deployed on a Field Programmable Gate Array (FPGA) that operated at reduced voltages and employed the proposed schemes. The SSL approach provides extremely low-overhead (≈0.2%) fault detection at the cost of lower error coverage and extra training, while ABFT incurs less than 8%overheads at run-time with close to 100% error detection rate. By using the solutions, substantial energy savings, i.e., up to 40.3%,without compromising the accuracy of the model was achieved

scholarly journals Low-Voltage Energy Efficient Neural Inference by Leveraging Fault Detection Techniques

2021 ◽  
Author(s):  
Mehdi Safarpour ◽  
Tommy Z. Deng ◽  
John Massingham ◽  
Lei Xun ◽  
Mohammad Sabokrou ◽  
...  
Keyword(s):  
Fault Detection ◽  
Error Detection ◽  
Energy Savings ◽  
Low Voltage ◽  
Building Blocks ◽  
Feedback Mechanism ◽  
Total System ◽  
Data Set ◽  
Detection Techniques ◽  
Energy Efficiency Improvement

Operating at reduced voltages offers substantial energy efficiency improvement but at the expense of increasing the probability of computational errors due to hardware faults. In this context, we targeted Deep Neural Networks (DNN) as emerging energy hungry building blocks in embedded applications. Without an error feedback mechanism, blind voltage down-scaling will result in degraded accuracy or total system failure. To enable safe voltage down-scaling, in this paper two solutions based on Self-Supervised Learning (SSL) and Algorithm Based Fault Tolerance (ABFT) were developed. A DNN model trained on MNIST data-set was deployed on a Field Programmable Gate Array (FPGA) that operated at reduced voltages and employed the proposed schemes. The SSL approach provides extremely low-overhead (≈0.2%) fault detection at the cost of lower error coverage and extra training, while ABFT incurs less than 8%overheads at run-time with close to 100% error detection rate. By using the solutions, substantial energy savings, i.e., up to 40.3%,without compromising the accuracy of the model was achieved

Assessing the Effect of Considering Multiple Data Properties on Detecting Potential Errors in Pavement Condition Data

2017 ◽  
Vol 2639 (1) ◽  
pp. 39-45
Author(s):  
Salar Zabihi Siabil ◽  
Nasir G. Gharaibeh
Keyword(s):  
Error Detection ◽  
False Negative ◽  
Data Set ◽  
Pavement Condition ◽  
Detection Techniques ◽  
Multiple Data ◽  
Texas Department ◽  
Maintenance And Rehabilitation ◽  
Current Error

The quality of pavement condition data can affect the assessment of current condition, predictions of future condition, and the reliability of maintenance and rehabilitation plans and funding need estimates at the network level. Thus, improving the quality of pavement condition data is an ongoing process for transportation agencies. Detecting potential errors in network-level pavement condition data is a primary step in assessing and enhancing the accuracy of this data. Current error detection techniques tend to focus on analyzing time series trends in pavement condition to identify unexpected changes that may denote data errors. However, there are additional properties of these data that can be used to identify potential errors, including variability within uniform performance families and consistency between multiple performance indicators. This paper assesses the effect of considering those data properties on detecting potential errors in pavement condition data. Three case analyses were defined such that each considered a different combination of these properties to identify likely errors. The analyses were performed on a pavement condition data set representing the Brownwood District roadway network of the Texas Department of Transportation. The results of this investigation indicate that considering such properties in a combined manner can reduce the numbers of false positive errors and false negative errors.

scholarly journals Low-Voltage Energy Efficient Neural Inference by Leveraging Fault Detection Techniques

2021 ◽  
Author(s):  
Mehdi Safarpour ◽  
Tommy Z. Deng ◽  
John Massingham ◽  
Lei Xun ◽  
Mohammad Sabokrou ◽  
...  
Keyword(s):  
Fault Detection ◽  
Energy Efficient ◽  
Low Voltage ◽  
Detection Techniques

Real-time error detection techniques based on FPGA

2013 ◽  
Vol 33 (5) ◽  
pp. 1459-1462
Author(s):  
Xiaoming JU ◽  
Jiehao ZHANG ◽  
Yizhong ZHANG
Keyword(s):  
Real Time ◽  
Error Detection ◽  
Time Error ◽  
Detection Techniques

scholarly journals Nanopore sequencing of single-cell transcriptomes with scCOLOR-seq

2021 ◽  
Author(s):  
Martin Philpott ◽  
Jonathan Watson ◽  
Anjan Thakurta ◽  
Tom Brown ◽  
Tom Brown ◽  
...  
Keyword(s):  
Single Cell ◽  
Error Detection ◽  
Single Cells ◽  
Fusion Transcript ◽  
Building Blocks ◽  
Myeloma Cell ◽  
Nanopore Sequencing ◽  
Long Read ◽  
Unique Molecular Identifier ◽  
Transcript Detection

AbstractHere we describe single-cell corrected long-read sequencing (scCOLOR-seq), which enables error correction of barcode and unique molecular identifier oligonucleotide sequences and permits standalone cDNA nanopore sequencing of single cells. Barcodes and unique molecular identifiers are synthesized using dimeric nucleotide building blocks that allow error detection. We illustrate the use of the method for evaluating barcode assignment accuracy, differential isoform usage in myeloma cell lines, and fusion transcript detection in a sarcoma cell line.

scholarly journals Data Augmentation Using Generative Adversarial Network for Automatic Machine Fault Detection Based on Vibration Signals

2021 ◽  
Vol 11 (5) ◽  
pp. 2166
Author(s):  
Van Bui ◽  
Tung Lam Pham ◽  
Huy Nguyen ◽  
Yeong Min Jang
Keyword(s):  
Fault Detection ◽  
Data Augmentation ◽  
Model Performance ◽  
Original Data ◽  
Fault Classification ◽  
Training Process ◽  
Generative Adversarial Network ◽  
Data Set ◽  
Adversarial Network ◽  
Machine Fault

In the last decade, predictive maintenance has attracted a lot of attention in industrial factories because of its wide use of the Internet of Things and artificial intelligence algorithms for data management. However, in the early phases where the abnormal and faulty machines rarely appeared in factories, there were limited sets of machine fault samples. With limited fault samples, it is difficult to perform a training process for fault classification due to the imbalance of input data. Therefore, data augmentation was required to increase the accuracy of the learning model. However, there were limited methods to generate and evaluate the data applied for data analysis. In this paper, we introduce a method of using the generative adversarial network as the fault signal augmentation method to enrich the dataset. The enhanced data set could increase the accuracy of the machine fault detection model in the training process. We also performed fault detection using a variety of preprocessing approaches and classified the models to evaluate the similarities between the generated data and authentic data. The generated fault data has high similarity with the original data and it significantly improves the accuracy of the model. The accuracy of fault machine detection reaches 99.41% with 20% original fault machine data set and 93.1% with 0% original fault machine data set (only use generate data only). Based on this, we concluded that the generated data could be used to mix with original data and improve the model performance.

Sign in / Sign up

Export Citation Format

Share Document