Noise Immunity of Oscillatory Computing Devices

Csaba, G.; Porod, Wolfgang

doi:10.1109/jxcdc.2020.3046558

Cited by 5 publications

(4 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In a previous work [32], Csaba and Porod highlighted the ONN robustness to electronic noise and have shown that ONN can tolerate a smaller SNR compared to amplitudebased computing systems achieving the same functionality. We express the oscillator SNR as the ratio between the peakto-peak voltage amplitude over the thermal noise standard deviation:…”

Section: Onn Size and Noise Limitationmentioning

confidence: 99%

“…9 shows the maximum fully-connected ONN size N max with respect to the oscillation frequency for various amplitudes. We consider SNR=3.5 as minimum achievable SNR that has been reported [32] in the case of two coupled-ring oscillators. With ∆V max =21 mV, we observe that the synaptic thermal noise limits the number of oscillators to N max =300 for f osc =10 MHz.…”

Section: Onn Size and Noise Limitationmentioning

confidence: 99%

“…oscillation frequency has to be reduced and the amplitude should increase. It is worthwhile to note that the minimum SNR might also depend on the ONN size, as Csaba and Porod [32] reported correct functionality for 100 coupled oscillators even for SNR<1. In literature, coupling oscillators has been shown efficient to reduce the phase noise [33], but yet little is known on the impact of noise on the oscillator synchronization and scaling of phase-based computing systems.…”

Section: Onn Size and Noise Limitationmentioning

confidence: 99%

See 2 more Smart Citations

Energy-Performance Assessment of Oscillatory Neural Networks Based on VO₂ Devices for Future Edge AI Computing

Delacour

Carapezzi

Abernot

et al. 2024

IEEE Trans. Neural Netw. Learning Syst.

View full text Add to dashboard Cite

Oscillatory Neural Network (ONN) is an emerging neuromorphic architecture composed of oscillators that implement neurons and coupled by synapses. ONNs exhibit rich dynamics and associative properties, which can be used to solve problems in the analog domain according to the paradigm let physics compute. For example, compact oscillators made of VO 2 material are good candidates for building low-power ONN architectures dedicated to AI applications at the edge like pattern recognition. However, little is known about the ONN scalability and its performances when implemented in hardware. Before deploying ONN, it is necessary to assess its computation time, energy consumption, performance and accuracy for a given application. Here, we consider a VO 2 -oscillator as an ONN building block and we perform circuit-level simulations to evaluate the ONN performances at the architecture level. Notably, we investigate how ONN computation time, energy and memory capacity scale with the number of oscillators. We show that ONN energy grows linearly when scaling up the network, making it suitable for large-scale integration at the edge. Furthermore, we investigate the design knobs for minimizing the ONN energy. Assisted by TCAD simulations, we report on scaling of VO 2 devices in crossbar geometry to decrease the oscillator voltage and energy. We benchmark ONN versus state-of-the-art architectures and observe that the ONN paradigm is a competitive energy efficient solution for scaled VO 2 devices. Finally, we present how ONN can efficiently detect edges in images captured on low-power edge devices.

show abstract

Section: Onn Size and Noise Limitationmentioning

confidence: 99%

Section: Onn Size and Noise Limitationmentioning

confidence: 99%

Section: Onn Size and Noise Limitationmentioning

confidence: 99%

See 1 more Smart Citation

Energy-Performance Assessment of Oscillatory Neural Networks Based on VO₂ Devices for Future Edge AI Computing

Delacour

Carapezzi

Abernot

et al. 2024

IEEE Trans. Neural Netw. Learning Syst.

View full text Add to dashboard Cite

show abstract

“…( Bottom ) If phases correspond to pixels of a grayscale image, the phase dynamics may be used to converge to predefined patterns (Csaba et al, 2016 ). The illustrations of convergence to ‘A' are taken from Csaba and Porod ( 2020b ).…”

Section: Methodsmentioning

confidence: 99%

Design of oscillatory neural networks by machine learning

Rudner,

Porod,

Csaba

2024

Front. Neurosci.

Self Cite

View full text Add to dashboard Cite

We demonstrate the utility of machine learning algorithms for the design of oscillatory neural networks (ONNs). After constructing a circuit model of the oscillators in a machine-learning-enabled simulator and performing Backpropagation through time (BPTT) for determining the coupling resistances between the ring oscillators, we demonstrate the design of associative memories and multi-layered ONN classifiers. The machine-learning-designed ONNs show superior performance compared to other design methods (such as Hebbian learning), and they also enable significant simplifications in the circuit topology. We also demonstrate the design of multi-layered ONNs that show superior performance compared to single-layer ones. We argue that machine learning can be a valuable tool to unlock the true computing potential of ONNs hardware.

show abstract

Creating electronic oscillator-based Ising machines without external injection locking

Vaidya

Kanthi

Shukla

2022

Sci Rep

View full text Add to dashboard Cite

Coupled electronic oscillators have recently been explored as a compact, integrated circuit- and room temperature operation-compatible hardware platform to design Ising machines. However, such implementations presently require the injection of an externally generated second-harmonic signal to impose the phase bipartition among the oscillators. In this work, we experimentally demonstrate a new electronic autaptic oscillator (EAO) that uses engineered feedback to eliminate the need for the generation and injection of the external second harmonic signal to minimize the Ising Hamiltonian. Unlike conventional relaxation oscillators that typically decay with a single time constant, the feedback in the EAO is engineered to generate two decay time constants which effectively helps generate the second harmonic signal internally. Using this oscillator design, we show experimentally, that a system of capacitively coupled EAOs exhibits the desired bipartition in the oscillator phases without the need for any external second harmonic injection, and subsequently, demonstrate its application in solving the computationally hard Maximum Cut (MaxCut) problem. Our work not only establishes a new oscillator design aligned to the needs of the oscillator Ising machine but also advances the efforts to creating application specific analog computing platforms.

show abstract

Noise Immunity of Oscillatory Computing Devices

Cited by 5 publications

References 23 publications

Energy-Performance Assessment of Oscillatory Neural Networks Based on VO₂ Devices for Future Edge AI Computing

Energy-Performance Assessment of Oscillatory Neural Networks Based on VO₂ Devices for Future Edge AI Computing

Design of oscillatory neural networks by machine learning

Creating electronic oscillator-based Ising machines without external injection locking

Contact Info

Product

Resources

About

Noise Immunity of Oscillatory Computing Devices

Cited by 5 publications

References 23 publications

Energy-Performance Assessment of Oscillatory Neural Networks Based on VO2 Devices for Future Edge AI Computing

Energy-Performance Assessment of Oscillatory Neural Networks Based on VO2 Devices for Future Edge AI Computing

Design of oscillatory neural networks by machine learning

Creating electronic oscillator-based Ising machines without external injection locking

Contact Info

Product

Resources

About

Energy-Performance Assessment of Oscillatory Neural Networks Based on VO₂ Devices for Future Edge AI Computing

Energy-Performance Assessment of Oscillatory Neural Networks Based on VO₂ Devices for Future Edge AI Computing