Introduction to Noise-Resilient Computing

Yanushkevich, Svetlana; Kasai, Seiya; Tangim, G.; Tran, A.H.; Mohamed, T.; Shmerko, Vlad

doi:10.2200/s00472ed1v01y201301dcs040

Cited by 4 publications

(4 citation statements)

References 130 publications

(228 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The "stress" or "no-stress" (0 or 1) state can be represented by their probabilities, e.g., the stress in one combatant is detected with a probability 0.7, and 0.8 in the second one. This area is called a "probabilistic logic computing" [73]. − A propagation mechanism.…”

Section: Computational Logicmentioning

confidence: 99%

Stress Propagation in Human-Robot Teams Based on Computational Logic Model

Shmerko¹,

Iwashita²,

Stoica³

et al. 2022

Preprint

View full text Add to dashboard Cite

Mission teams are exposed to the emotional toll of life and death decisions. These are small groups of specially trained people supported by intelligent machines for dealing with stressful environments and scenarios. We developed a composite model for stress monitoring in such teams of human and autonomous machines. This modelling aims to identify the conditions that may contribute to mission failure. The proposed model is composed of three parts: 1) a computational logic part that statically describes the stress states of teammates; 2) a decision part that manifests the mission status at any time; 3) a stress propagation part based on standard Susceptible-Infected-Susceptible (SIS) paradigm. In contrast to the approaches such as agent-based, random-walk and game models, the proposed model combines various mechanisms to satisfy the conditions of stress propagation in small groups. Our core apporach involves data structures such as decision tables and decision diagrams. These tools are adaptable to human-machine teaming as well.

show abstract

Section: Computational Logicmentioning

confidence: 99%

Stress Propagation in Human-Robot Teams Based on Computational Logic Model

Shmerko¹,

Iwashita²,

Stoica³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…[23], [24]. Even for specialized nanoscale circuits, the International Technology Roadmap for semiconductors predicts that as devices become smaller, thermal noise itself will cause systems to fail catastrophically during normal operation even without supply voltage scaling [7], thus increasing the need for fault-tolerant training.…”

Section: A Related Workmentioning

confidence: 99%

“…It is often speculated [4] that the error-prone nature of brain's hardware helps it be more efficient: rather than paying the cost at a component level, it may be more efficient to accept component-level errors, and utilize sophisticated error-correction mechanisms for overall reliability of the computation 1 . The brain operates at a surprisingly low power of about 15 W [7], and attains high accuracy and speeds, despite individual neurons in the brain being slow and error-prone [4], [5], [8]- [10].…”

Section: Introductionmentioning

confidence: 99%

CodeNet: Training Large Scale Neural Networks in Presence of Soft-Errors

Dutta¹,

Bai²,

Low³

et al. 2019

Preprint

View full text Add to dashboard Cite

This work proposes the first strategy to make distributed training of neural networks resilient to computing errors, a problem that has remained unsolved despite being first posed in 1956 by von Neumann. He also speculated that the efficiency and reliability of the human brain is obtained by allowing for low power but error-prone components with redundancy for error-resilience. It is surprising that this problem remains open, even as massive artificial neural networks are being trained on increasingly low-cost and unreliable processing units. Our coding-theory-inspired strategy, "CodeNet," solves this problem by addressing three challenges in the science of reliable computing: (i) Providing the first strategy for error-resilient neural network training by encoding each layer separately; (ii) Keeping the overheads of coding (encoding/error-detection/decoding) low by obviating the need to re-encode the updated parameter matrices after each iteration from scratch. (iii) Providing a completely decentralized implementation with no central node (which is a single point of failure), allowing all primary computational steps to be error-prone. We theoretically demonstrate that CodeNet has higher error tolerance than replication, which we leverage to speedup computation time. Simultaneously, CodeNet requires lower redundancy than replication, and equal computational and communication costs in scaling sense. We first theoretically demonstrate the benefits of using CodeNet over replication in reducing the expected computation time accounting for checkpointing. Our experiments show that CodeNet achieves the best accuracy-runtime tradeoff compared to both replication and uncoded strategies. CodeNet is a significant step towards biologically plausible neural network training, that could hold the key to orders of magnitude efficiency improvements.

show abstract

“…18 Considering that our system always reaches one of the solutions except deadlock, this suggests the capability of spontaneous searching of more than one solution. In the physical electronic circuit, fluctuation is provided from the external electromagnetic noise, low frequency noise in FETs, 19 statistical variation in nanodevices, 20 and so on. In addition, the degree of the parameter variation in our system affected the solution space that the system could search.…”

mentioning

confidence: 99%

Amoeba-inspired computing architecture implemented using charge dynamics in parallel capacitance network

Kasai

Aono

Naruse

2013

Applied Physics Letters

View full text Add to dashboard Cite

We propose an electronic system for implementing a biologically inspired computing architecture, called "amoeba-inspired computing," for solving computationally demanding problems. The system consists of a parallel capacitance network. The spatiotemporal dynamics of an amoeboid organism exhibiting the sophisticated ability of exploring a solution space is mimicked using dynamics in charging the capacitors under charge conservation. The system for solving an instance of a four-variable constraint satisfaction problem (CSP) is implemented using an electronic circuit simulator, which successfully finds solutions. We also found that small fluctuations inherently involved in electronic devices can be used to explore solution space. (C) 2013 AIP Publishing LLC

show abstract

Introduction to Noise-Resilient Computing

Cited by 4 publications

References 130 publications

Stress Propagation in Human-Robot Teams Based on Computational Logic Model

Stress Propagation in Human-Robot Teams Based on Computational Logic Model

CodeNet: Training Large Scale Neural Networks in Presence of Soft-Errors

Amoeba-inspired computing architecture implemented using charge dynamics in parallel capacitance network

Contact Info

Product

Resources

About