Abstract
Neural Network (NN) accelerators are essential in many emerging applications e.g., autonomous systems in making mission-critical decisions, health-care solutions to assist with diagnoses, etc. Any soft or hard failure during operation can potentially have catastrophic consequences in many of these applications. For instance, inaccurate classification during object recognition and tracking in autonomous vehicles can lead to crashes and subsequent injuries to the passengers 8. Therefore, testing Neural Network accelerators to ensure reliability and integrity of the underlying hardware is a crucial task to ensure the functionality, especially the ones that are used in mission-critical applications. Conventional functional, stuck-at and delay tests are not sufficient to characterize the ML systems since they face new test and validation challenges. This paper is aimed to provide a perspective on new test requirements and design for test techniques to cover ML features and detect various type of faults in NN accelerator. We discuss First-In-First-Out (FIFO) and Scratchpad based neural network hardware accelerators and propose test methods to detect the faults as well as fault location in different modules of the accelerator including MAC unit, Activation function module, and Processing Element (PE) registers.
| Original language | English |
|---|---|
| Journal | IEEE Journal on Emerging and Selected Topics in Circuits and Systems |
| DOIs | |
| Publication status | Published - 2019 Jan 1 |
Fingerprint
Keywords
- Artificial neural networks
- Biological neural networks
- Circuit faults
- DFT
- Fault tolerance
- Fault tolerant systems
- Hardware
- Hardware Accelerator
- Neural Network
- Neurons
- Stuck at Fault
ASJC Scopus subject areas
- Electrical and Electronic Engineering
Cite this
A Perspective on Test Methodologies for Supervised Machine Learning Accelerators. / Motaman, Seyedhamidreza; Ghosh, Swaroop; Park, Jongsun.
In: IEEE Journal on Emerging and Selected Topics in Circuits and Systems, 01.01.2019.Research output: Contribution to journal › Article
}
TY - JOUR
T1 - A Perspective on Test Methodologies for Supervised Machine Learning Accelerators
AU - Motaman, Seyedhamidreza
AU - Ghosh, Swaroop
AU - Park, Jongsun
PY - 2019/1/1
Y1 - 2019/1/1
N2 - Neural Network (NN) accelerators are essential in many emerging applications e.g., autonomous systems in making mission-critical decisions, health-care solutions to assist with diagnoses, etc. Any soft or hard failure during operation can potentially have catastrophic consequences in many of these applications. For instance, inaccurate classification during object recognition and tracking in autonomous vehicles can lead to crashes and subsequent injuries to the passengers 8. Therefore, testing Neural Network accelerators to ensure reliability and integrity of the underlying hardware is a crucial task to ensure the functionality, especially the ones that are used in mission-critical applications. Conventional functional, stuck-at and delay tests are not sufficient to characterize the ML systems since they face new test and validation challenges. This paper is aimed to provide a perspective on new test requirements and design for test techniques to cover ML features and detect various type of faults in NN accelerator. We discuss First-In-First-Out (FIFO) and Scratchpad based neural network hardware accelerators and propose test methods to detect the faults as well as fault location in different modules of the accelerator including MAC unit, Activation function module, and Processing Element (PE) registers.
AB - Neural Network (NN) accelerators are essential in many emerging applications e.g., autonomous systems in making mission-critical decisions, health-care solutions to assist with diagnoses, etc. Any soft or hard failure during operation can potentially have catastrophic consequences in many of these applications. For instance, inaccurate classification during object recognition and tracking in autonomous vehicles can lead to crashes and subsequent injuries to the passengers 8. Therefore, testing Neural Network accelerators to ensure reliability and integrity of the underlying hardware is a crucial task to ensure the functionality, especially the ones that are used in mission-critical applications. Conventional functional, stuck-at and delay tests are not sufficient to characterize the ML systems since they face new test and validation challenges. This paper is aimed to provide a perspective on new test requirements and design for test techniques to cover ML features and detect various type of faults in NN accelerator. We discuss First-In-First-Out (FIFO) and Scratchpad based neural network hardware accelerators and propose test methods to detect the faults as well as fault location in different modules of the accelerator including MAC unit, Activation function module, and Processing Element (PE) registers.
KW - Artificial neural networks
KW - Biological neural networks
KW - Circuit faults
KW - DFT
KW - Fault tolerance
KW - Fault tolerant systems
KW - Hardware
KW - Hardware Accelerator
KW - Neural Network
KW - Neurons
KW - Stuck at Fault
UR - http://www.scopus.com/inward/record.url?scp=85070720009&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85070720009&partnerID=8YFLogxK
U2 - 10.1109/JETCAS.2019.2933678
DO - 10.1109/JETCAS.2019.2933678
M3 - Article
AN - SCOPUS:85070720009
JO - IEEE Journal on Emerging and Selected Topics in Circuits and Systems
JF - IEEE Journal on Emerging and Selected Topics in Circuits and Systems
SN - 2156-3357
ER -