WILLIAMS Yohanna Yerima
をテンプレートにして作成
[
トップ
] [
新規
|
一覧
|
単語検索
|
最終更新
|
ヘルプ
|
ログイン
]
開始行:
[[Members-Internal]]
//CENTER:COLOR(#81114F){SIZE(30){''Adaptive Robotic Arm/Prosthetic Control based on NASH''}}
CENTER:COLOR(#81111F){SIZE(30){''Fault-tolerant Routing Algorithm and Architecture for Reconfigurable 3D-NoC-based Neuromorphic Systems''}}
*Background [#d3104867]
The basic processing units in the brain are neurons and synapses that are interconnected in a rather parallel pattern with information processing capabilities. The attempt to mimic and build neural systems in hardware devices to emulate the key information processing principles of the brain is implemented on a massively parallel computing system with thousands of neurons. The neuron (nodes) interconnection within these hardwares plays a key role in their performance but with the high number of components significantly increasing through scalability, the probability of failure also increases. SNN that communicates via spikes in an asynchronous fashion has been used of late to process information (packets) between interconnected neurons via a synapse in neuromorphic hardwares however for large-scale spiking neural networks, the conventional spike-based hardware could encounter the challenge of inefficient spikes distribution and partial or total failure due to inflexibility in routing strategy and neuron mapping on its architecture that could render some part of the hardware isolated from the rest and severe power consumption. Routing and mapping strategies reconfigured in neuromorphic hardwares are designed to yield robust hardware with reduced power dissipation, increased throughput, and minimal latency for large-scale computing applications.
*Motivation [#w893ae2e]
Spiking neural networks (SNNs) mimic the information processing pattern in the mammalian brain based on parallel arrays of neurons that communicate via spike events. To explore their energy efficiency and computational speed-up characteristics, they are implemented in hardware (Neuromorphic) with NoC (network-on-chip) for scalable communication and an effective packet distribution strategy with high bandwidth. However, one of the main problems of these (SNNs) implementations is their reliability potential despite several claims that the model has some intrinsic fault-tolerance properties. For large-scale networks with a dense population of spiking neurons, the fault risk becomes more due to the accumulation of probable occurrence of faults in conventional hardwares. Also, an inefficient packet routing strategy may lead to latency build-up, deadlocks, and several dead ends in the network. This demonstrates that reconfiguring the neuromorphic hardwares routing strategy and equipping it with fault detection and tolerant mechanism would be crucial in mitigating the risk of failures and improving the overall system output performance.
*Goal [#v20a6f9f]
CENTER:&ref(fault.png,,60%);
*Research Schedule [#md706e02]
|Date|Task|
|☑️May 19th - May 30, 2022| SNN architecture and algorithm study |
|☑️ May 31th - June 04, 2022| Study fault types and positions of occurrence |
|☑️June 05 - June 08, 2022| Study fault-tolerant routing architectures |
|☑️ June 09 - June 19, 2020| Revisit the proposed new strategy for routing and mapping (focusing router) |
|☑️June 20- July 01, 2022| Run SNPC codes and evaluate Area, Power consumption|
|☑️ July 02 - July 12, 2022| Run NASH codes and evaluate accuracy on MNIST|
|☑️July 13 - July 15, 2022| Run MigSpike codes|
|☑️ July 16 - July 22, 2022| Train MNIST using STDP|
|☑️ July 23 - July 28, 2022| Run the faults and dropout/drop connects on|
|☑️July 29- August 10, 2022| Propose a better approach to neuron dropout/drop connect |
|☑️ August 11 - August 15, 2022| Explore how the mechanism can be implemented in hardware (NASH) to solve the Migspike drawback|
|☑️ August 16 - August 18, 2022| Run the faults and dropout/drop connects on the trained SNN model|
|☑️ August 17 - August 25, 2022| Reconsider to drop out neurons instead of weight using pruning|
|☑️August 26 - August 30, 2022| Propose an approach to dropout (non-contributing) neurons using pruning|
|☑️ August 31 - September 10, 2022| Propose an approach to reduce faulty neurons in Mig-NASH for remapping using MigSpike|
|☑️Septmber 11 - September 18, 2022| Revisit the approach for final implementation|
|☑️ September 19 - September 30, 2022| Implement the proposed pruning method and perform evaluation (accuracy, inference time) and summarize the result with figures and tables|
|☑️ October 01 - October 05, 2022| Prepare and present part of research work at the PG forum UoA.|
|☑️ October 06- October 08, 2022|Update overleaf project for my research|
|☑️October 09- October 24, 2022| Continue implementation evaluation and explore other works to make comparison|
|☑️October 25- October 31, 2022| Begin initial Journal and ETLTC 2023 paper draft(s)|
|☑️ November 01- November 10, 2022|Implement proposed Fault recovery approach: Software|
|☑️ November 11- November 30, 2022| Implement SNN models with the fault recovery method on NASH and later in the MIGSPIKE ALOGIRTHM and perform evaluation (Energy consumption, Accuracy)|
|☑️ December 01- December 02, 2022| Study the updated MIGSPIKE CODES|
|☑️ December 03- December 05, 2022| Update (modify the abstract, include all the results from the implemented recovery method) the IEEE Conference Paper (Final) and the Journal draft (Second)|
|☑️ December 06- December 16, 2022| Continue with the implementation of the proposed recovery method in NASH|
|☑️ December 17- December 20, 2022| Update the Journal draft paper|
|☑️ December 21- December 30, 2022| Propose an optimization method for FT during mapping|
|☑️ December 31- January 07, 2023| Update the journal draft paper|
|☑️ January 08- January 13, 2023| Revisit and simulate the proposed optimization method|
|☑️ January 14- January 18, 2023| Update the journal draft paper|
|☑️ January 19- January 25, 2023| Integrate the proposed algorithm(s) into the mapping algorithm,and perform evaluation|
|☑️ January 26- January 31, 2023| Make a final update on journal draft paper|
|☑️ February 01- February 18, 2023| Proofread the draft and submit to other co-authors for review and comments|
|☑️ February 19- February 28, 2023| Revise and update the journal after other co-authors review|
|☑️ March 01- March 23, 2023| Submit the revised draft paper for second co-authors review and comments|
|☑️ March 24- April 04, 2023| Revise the draft paper after second other co-authors' comments and review|
|☑️ April 05- April 14, 2023| Re-submit the revised and updated draft paper based on comments and review from other co-authors'|
|☑️ April 15- April 17, 2023| Proofread the paper and submit to IEEE Access.|
|☑️ April 18- April 30, 2023| Explore conducting experiments to improve the evaluation section in the submitted journal paper|
|☑️ May 1st - May 02, 2023| Survey on reconfigurable neuromorphic systems|
|☑️ May 03 - May 05, 2023| Survey on spiking neurons models/types|
|☑️ May 08 - May 12, 2023| Explore on Izhikevich spiking neurons and implement in the SNPC|
|☑️ May 13 - May 16, 2023| Summarize the state-of-the-art mapping methods on the NASH and identify their limitation|
|☑️ May 17 - May 19, 2023| Survey on the concept of Neural reuse|
|☑️ May 20 - May 29, 2023| Propose a robust mapping method to optimize the state of the FT mapping methods on the NASH and/or a new FT mapping method based on the concept of Neural reuse|
|☑️ May 31 - June 02, 2023|Begin an IEEE conference draft|
|☑️ June 03 - June 06, 2023| Propose an optimized fault-recovery method in SNNs|
|☑️ June 07 - June 10, 2023| Conduct experiments and evaluate the accuracy recovery of SNNs with the fault recovery method|
|☑️ June 11 - June 12, 2023| Prepare an abstract and submit to the conference chair|
|☑️ June 13 - June 20, 2023| Begin initial second journal draft|
|☑️ June 21 - June 25, 2023| Update the conference draft|
|☑️ June 26 - June 28, 2023| Review the proposed robust mapping method|
|☑️ June 29 - June 30, 2023| Review the IEEE conference draft and submit|
|☑️ July 01 - July 10, 2023| Conduct experiments for the proposed robust mapping method|
|☑️ July 11 - July 30, 2023| Update the second IEEE Access journal draft|
|☑️ August 01 - August 10, 2023| Submit to other co-authors, update accordingly after their review, and submit|
|☑️ August 11 - September 02, 2023| Begin thesis writing|
|☑️ September 03 - September 05, 2023| Make minor edits on the accepted IEEE Access journal paper and upload final files|
|☑️ September 06 - September 30, 2023| Continue with the thesis writing|
|☑️ October 01 - October 16, 2023| Prepare slides for preliminary review|
|☑️ October 17 - October 30, 2023| Prepare preliminary review report and submit to SAD|
|☑️ November 01 - November 10, 2023| Investigate on Visual Feedback in Prosthesis|
|☑️ November 11 - November 20, 2023| Print a new 3D prosthetic hand|
|☑️ November 21 - November 24, 2023| Implement vibrotactile feedback on the AIzuhand|
|☑️ November 25 - December 11, 2023| Survey on Feedback Systems in Prosthetic Control|
CENTER:Schedule Updated on November 27, 2023
*Doctoral Dissertation Review Procedure (AY2024 Spring [[Ref>https://www.u-aizu.ac.jp/en/graduate/curriculum/doctor/]]) [#cf0b2887]
|Date|Task|
|☑️May 15, 2023|Submission of dissertation title and list of referees|
|☑️Oct 06, 2023|Submission of the documents for the Doctoral Dissertation Preliminary Review|
|☑️Oct 17, 2023 Doctoral Dissertation Preliminary Review|
|☑️ Oct 27, 2023 Submission of Doctoral Dissertation Preliminary Review Report|
|☑️Nov 13, 2023 Doctoral Dissertation Final Review Decision Date|
|☑️Dec 15, 2023 Doctoral Dissertation Final Review Documents Submission|
|☑️Jan 09, 2023 Doctoral Dissertation Final Review|
CENTER:COLOR(green){'Schedule last Updated on: November 27, 2023'}
*Achievements [#g0c8f9db]
***Accepted Journal Papers [#p4f63a6e]
-"Fault-Tolerant Spiking Neural Network Mapping Algorithm and Architecture to 3D-NoC-Based Neuromorphic Systems," in IEEE Access, vol. 11, pp. 52429-52443, 2023, doi: 10.1109/ACCESS.2023.3278802 ([[Overleaf>https://drive.google.com/file/d/1KtWLwS_CPPluFGjaXsepQUnXgkEPU-AS/view?usp=sharing]],
[[pdf>https://drive.google.com/file/d/1RpHLcpUiS693jH7rBtYV5iDImOfZE4ob/view?usp=sharing]])
-"R-MaS3N: Robust Mapping of Spiking Neural Networks to 3D-NoC-Based Neuromorphic Systems for Enhanced Reliability in IEEE Access, vol. 11, pp. 94664-94678, 2023, doi: 10.1109/ACCESS.2023.3311031 ([[Overleaf>https://drive.google.com/file/d/1lAoku7dXklGf-MqJnumzRwrZaaBKosu6/view?usp=sharing]],
[[pdf>https://drive.google.com/file/d/1tVSWgju6Qqadv6cK4L-hGVWfW2nH1hH0/view?usp=sharing]])
***Accepted Conference Papers [#yc693d3d]
-@inproceedings{Williams2023fault,
title={Fault Recovery in Spiking Neural Networks Through Target and Selection of Faulty Neurons For 3D Spiking Neuromorphic Processors},
author={Williams Yohanna Yerima, Dang Khanh Nam, and Abdallah Abderazek Ben},
booktitle={6th IEEE International Conference on Knowledge Innovation and Invention 2023},
pages={-},
year={2023},
organization={IEEE},
}
([[Overleaf>https://drive.google.com/file/d/1zoE477Kx2pNaZF_8-Q7Nnaw9RW77HL5F/view?usp=sharing]],
[[pdf>https://drive.google.com/file/d/1CP3ErjBPltxutKGcltCK_g8PUWqhpk_r/view?usp=sharing]])
*My Shared GDrive [#pa27546c]
-[[My Shared GoogleDrive>https://drive.google.com/drive/folders/1t2e6GSNA0h43EItpI8Q2YGHsNd8JSTJ1?usp=sharing]]
-[[Ebooks>https://drive.google.com/drive/folders/1t0ewhEhe4Oa9AaTPKdmUfJib0a37JEdD?usp=sharing]]
***References [#f7a2c2f7]
-“Dropout and DropConnect for reliable neuromorphic inference under communication constraints in network connectivity,” IEEE Journal on Emerging and Selected Topics in Circuits and Systems, 2019
-“A majority-based reliability-aware task mapping in high-performance homogenous NoC architectures,” ACM Transactions on Embedded Computing Systems (TECS), 2107
-“Fault-tolerant systolic array based accelerators for deep neural network execution,” IEEE Design & Test, 2019.
-“Homeostatic fault tolerance in spiking neural networks: a dynamic hardware perspective,” IEEE Transactions on Circuits and Systems I: Regular Papers 2017
-“Fault-tolerant network-on-chip design with flexible spare core placement,”J. Emerg. Technol. Comput. Syst., 2019
-“Fault-tolerant spike routing algorithm and architecture for three dimensional NoC-based neuromorphic systems,”IEEE Access, 2019.
MigSpike: A Migration Based Algorithm and Architecture for Scalable Robust Neuromorphic Systems”, IEEE Transactions on Emerging Topics in Computing (TETC) 2022.
-----
終了行:
[[Members-Internal]]
//CENTER:COLOR(#81114F){SIZE(30){''Adaptive Robotic Arm/Prosthetic Control based on NASH''}}
CENTER:COLOR(#81111F){SIZE(30){''Fault-tolerant Routing Algorithm and Architecture for Reconfigurable 3D-NoC-based Neuromorphic Systems''}}
*Background [#d3104867]
The basic processing units in the brain are neurons and synapses that are interconnected in a rather parallel pattern with information processing capabilities. The attempt to mimic and build neural systems in hardware devices to emulate the key information processing principles of the brain is implemented on a massively parallel computing system with thousands of neurons. The neuron (nodes) interconnection within these hardwares plays a key role in their performance but with the high number of components significantly increasing through scalability, the probability of failure also increases. SNN that communicates via spikes in an asynchronous fashion has been used of late to process information (packets) between interconnected neurons via a synapse in neuromorphic hardwares however for large-scale spiking neural networks, the conventional spike-based hardware could encounter the challenge of inefficient spikes distribution and partial or total failure due to inflexibility in routing strategy and neuron mapping on its architecture that could render some part of the hardware isolated from the rest and severe power consumption. Routing and mapping strategies reconfigured in neuromorphic hardwares are designed to yield robust hardware with reduced power dissipation, increased throughput, and minimal latency for large-scale computing applications.
*Motivation [#w893ae2e]
Spiking neural networks (SNNs) mimic the information processing pattern in the mammalian brain based on parallel arrays of neurons that communicate via spike events. To explore their energy efficiency and computational speed-up characteristics, they are implemented in hardware (Neuromorphic) with NoC (network-on-chip) for scalable communication and an effective packet distribution strategy with high bandwidth. However, one of the main problems of these (SNNs) implementations is their reliability potential despite several claims that the model has some intrinsic fault-tolerance properties. For large-scale networks with a dense population of spiking neurons, the fault risk becomes more due to the accumulation of probable occurrence of faults in conventional hardwares. Also, an inefficient packet routing strategy may lead to latency build-up, deadlocks, and several dead ends in the network. This demonstrates that reconfiguring the neuromorphic hardwares routing strategy and equipping it with fault detection and tolerant mechanism would be crucial in mitigating the risk of failures and improving the overall system output performance.
*Goal [#v20a6f9f]
CENTER:&ref(fault.png,,60%);
*Research Schedule [#md706e02]
|Date|Task|
|☑️May 19th - May 30, 2022| SNN architecture and algorithm study |
|☑️ May 31th - June 04, 2022| Study fault types and positions of occurrence |
|☑️June 05 - June 08, 2022| Study fault-tolerant routing architectures |
|☑️ June 09 - June 19, 2020| Revisit the proposed new strategy for routing and mapping (focusing router) |
|☑️June 20- July 01, 2022| Run SNPC codes and evaluate Area, Power consumption|
|☑️ July 02 - July 12, 2022| Run NASH codes and evaluate accuracy on MNIST|
|☑️July 13 - July 15, 2022| Run MigSpike codes|
|☑️ July 16 - July 22, 2022| Train MNIST using STDP|
|☑️ July 23 - July 28, 2022| Run the faults and dropout/drop connects on|
|☑️July 29- August 10, 2022| Propose a better approach to neuron dropout/drop connect |
|☑️ August 11 - August 15, 2022| Explore how the mechanism can be implemented in hardware (NASH) to solve the Migspike drawback|
|☑️ August 16 - August 18, 2022| Run the faults and dropout/drop connects on the trained SNN model|
|☑️ August 17 - August 25, 2022| Reconsider to drop out neurons instead of weight using pruning|
|☑️August 26 - August 30, 2022| Propose an approach to dropout (non-contributing) neurons using pruning|
|☑️ August 31 - September 10, 2022| Propose an approach to reduce faulty neurons in Mig-NASH for remapping using MigSpike|
|☑️Septmber 11 - September 18, 2022| Revisit the approach for final implementation|
|☑️ September 19 - September 30, 2022| Implement the proposed pruning method and perform evaluation (accuracy, inference time) and summarize the result with figures and tables|
|☑️ October 01 - October 05, 2022| Prepare and present part of research work at the PG forum UoA.|
|☑️ October 06- October 08, 2022|Update overleaf project for my research|
|☑️October 09- October 24, 2022| Continue implementation evaluation and explore other works to make comparison|
|☑️October 25- October 31, 2022| Begin initial Journal and ETLTC 2023 paper draft(s)|
|☑️ November 01- November 10, 2022|Implement proposed Fault recovery approach: Software|
|☑️ November 11- November 30, 2022| Implement SNN models with the fault recovery method on NASH and later in the MIGSPIKE ALOGIRTHM and perform evaluation (Energy consumption, Accuracy)|
|☑️ December 01- December 02, 2022| Study the updated MIGSPIKE CODES|
|☑️ December 03- December 05, 2022| Update (modify the abstract, include all the results from the implemented recovery method) the IEEE Conference Paper (Final) and the Journal draft (Second)|
|☑️ December 06- December 16, 2022| Continue with the implementation of the proposed recovery method in NASH|
|☑️ December 17- December 20, 2022| Update the Journal draft paper|
|☑️ December 21- December 30, 2022| Propose an optimization method for FT during mapping|
|☑️ December 31- January 07, 2023| Update the journal draft paper|
|☑️ January 08- January 13, 2023| Revisit and simulate the proposed optimization method|
|☑️ January 14- January 18, 2023| Update the journal draft paper|
|☑️ January 19- January 25, 2023| Integrate the proposed algorithm(s) into the mapping algorithm,and perform evaluation|
|☑️ January 26- January 31, 2023| Make a final update on journal draft paper|
|☑️ February 01- February 18, 2023| Proofread the draft and submit to other co-authors for review and comments|
|☑️ February 19- February 28, 2023| Revise and update the journal after other co-authors review|
|☑️ March 01- March 23, 2023| Submit the revised draft paper for second co-authors review and comments|
|☑️ March 24- April 04, 2023| Revise the draft paper after second other co-authors' comments and review|
|☑️ April 05- April 14, 2023| Re-submit the revised and updated draft paper based on comments and review from other co-authors'|
|☑️ April 15- April 17, 2023| Proofread the paper and submit to IEEE Access.|
|☑️ April 18- April 30, 2023| Explore conducting experiments to improve the evaluation section in the submitted journal paper|
|☑️ May 1st - May 02, 2023| Survey on reconfigurable neuromorphic systems|
|☑️ May 03 - May 05, 2023| Survey on spiking neurons models/types|
|☑️ May 08 - May 12, 2023| Explore on Izhikevich spiking neurons and implement in the SNPC|
|☑️ May 13 - May 16, 2023| Summarize the state-of-the-art mapping methods on the NASH and identify their limitation|
|☑️ May 17 - May 19, 2023| Survey on the concept of Neural reuse|
|☑️ May 20 - May 29, 2023| Propose a robust mapping method to optimize the state of the FT mapping methods on the NASH and/or a new FT mapping method based on the concept of Neural reuse|
|☑️ May 31 - June 02, 2023|Begin an IEEE conference draft|
|☑️ June 03 - June 06, 2023| Propose an optimized fault-recovery method in SNNs|
|☑️ June 07 - June 10, 2023| Conduct experiments and evaluate the accuracy recovery of SNNs with the fault recovery method|
|☑️ June 11 - June 12, 2023| Prepare an abstract and submit to the conference chair|
|☑️ June 13 - June 20, 2023| Begin initial second journal draft|
|☑️ June 21 - June 25, 2023| Update the conference draft|
|☑️ June 26 - June 28, 2023| Review the proposed robust mapping method|
|☑️ June 29 - June 30, 2023| Review the IEEE conference draft and submit|
|☑️ July 01 - July 10, 2023| Conduct experiments for the proposed robust mapping method|
|☑️ July 11 - July 30, 2023| Update the second IEEE Access journal draft|
|☑️ August 01 - August 10, 2023| Submit to other co-authors, update accordingly after their review, and submit|
|☑️ August 11 - September 02, 2023| Begin thesis writing|
|☑️ September 03 - September 05, 2023| Make minor edits on the accepted IEEE Access journal paper and upload final files|
|☑️ September 06 - September 30, 2023| Continue with the thesis writing|
|☑️ October 01 - October 16, 2023| Prepare slides for preliminary review|
|☑️ October 17 - October 30, 2023| Prepare preliminary review report and submit to SAD|
|☑️ November 01 - November 10, 2023| Investigate on Visual Feedback in Prosthesis|
|☑️ November 11 - November 20, 2023| Print a new 3D prosthetic hand|
|☑️ November 21 - November 24, 2023| Implement vibrotactile feedback on the AIzuhand|
|☑️ November 25 - December 11, 2023| Survey on Feedback Systems in Prosthetic Control|
CENTER:Schedule Updated on November 27, 2023
*Doctoral Dissertation Review Procedure (AY2024 Spring [[Ref>https://www.u-aizu.ac.jp/en/graduate/curriculum/doctor/]]) [#cf0b2887]
|Date|Task|
|☑️May 15, 2023|Submission of dissertation title and list of referees|
|☑️Oct 06, 2023|Submission of the documents for the Doctoral Dissertation Preliminary Review|
|☑️Oct 17, 2023 Doctoral Dissertation Preliminary Review|
|☑️ Oct 27, 2023 Submission of Doctoral Dissertation Preliminary Review Report|
|☑️Nov 13, 2023 Doctoral Dissertation Final Review Decision Date|
|☑️Dec 15, 2023 Doctoral Dissertation Final Review Documents Submission|
|☑️Jan 09, 2023 Doctoral Dissertation Final Review|
CENTER:COLOR(green){'Schedule last Updated on: November 27, 2023'}
*Achievements [#g0c8f9db]
***Accepted Journal Papers [#p4f63a6e]
-"Fault-Tolerant Spiking Neural Network Mapping Algorithm and Architecture to 3D-NoC-Based Neuromorphic Systems," in IEEE Access, vol. 11, pp. 52429-52443, 2023, doi: 10.1109/ACCESS.2023.3278802 ([[Overleaf>https://drive.google.com/file/d/1KtWLwS_CPPluFGjaXsepQUnXgkEPU-AS/view?usp=sharing]],
[[pdf>https://drive.google.com/file/d/1RpHLcpUiS693jH7rBtYV5iDImOfZE4ob/view?usp=sharing]])
-"R-MaS3N: Robust Mapping of Spiking Neural Networks to 3D-NoC-Based Neuromorphic Systems for Enhanced Reliability in IEEE Access, vol. 11, pp. 94664-94678, 2023, doi: 10.1109/ACCESS.2023.3311031 ([[Overleaf>https://drive.google.com/file/d/1lAoku7dXklGf-MqJnumzRwrZaaBKosu6/view?usp=sharing]],
[[pdf>https://drive.google.com/file/d/1tVSWgju6Qqadv6cK4L-hGVWfW2nH1hH0/view?usp=sharing]])
***Accepted Conference Papers [#yc693d3d]
-@inproceedings{Williams2023fault,
title={Fault Recovery in Spiking Neural Networks Through Target and Selection of Faulty Neurons For 3D Spiking Neuromorphic Processors},
author={Williams Yohanna Yerima, Dang Khanh Nam, and Abdallah Abderazek Ben},
booktitle={6th IEEE International Conference on Knowledge Innovation and Invention 2023},
pages={-},
year={2023},
organization={IEEE},
}
([[Overleaf>https://drive.google.com/file/d/1zoE477Kx2pNaZF_8-Q7Nnaw9RW77HL5F/view?usp=sharing]],
[[pdf>https://drive.google.com/file/d/1CP3ErjBPltxutKGcltCK_g8PUWqhpk_r/view?usp=sharing]])
*My Shared GDrive [#pa27546c]
-[[My Shared GoogleDrive>https://drive.google.com/drive/folders/1t2e6GSNA0h43EItpI8Q2YGHsNd8JSTJ1?usp=sharing]]
-[[Ebooks>https://drive.google.com/drive/folders/1t0ewhEhe4Oa9AaTPKdmUfJib0a37JEdD?usp=sharing]]
***References [#f7a2c2f7]
-“Dropout and DropConnect for reliable neuromorphic inference under communication constraints in network connectivity,” IEEE Journal on Emerging and Selected Topics in Circuits and Systems, 2019
-“A majority-based reliability-aware task mapping in high-performance homogenous NoC architectures,” ACM Transactions on Embedded Computing Systems (TECS), 2107
-“Fault-tolerant systolic array based accelerators for deep neural network execution,” IEEE Design & Test, 2019.
-“Homeostatic fault tolerance in spiking neural networks: a dynamic hardware perspective,” IEEE Transactions on Circuits and Systems I: Regular Papers 2017
-“Fault-tolerant network-on-chip design with flexible spare core placement,”J. Emerg. Technol. Comput. Syst., 2019
-“Fault-tolerant spike routing algorithm and architecture for three dimensional NoC-based neuromorphic systems,”IEEE Access, 2019.
MigSpike: A Migration Based Algorithm and Architecture for Scalable Robust Neuromorphic Systems”, IEEE Transactions on Emerging Topics in Computing (TETC) 2022.
-----
ページ名: