Effect of weight overlap region on neuromorphic system with memristive synaptic devices

Lee, Geun Ho; Kim, Tae-Hyeon; Song, Min; Park, Jin-Woo; Kim, Sung-Joon; Hong, Kyungho; Kim, Yoon; Park, Byung‐Gook; Kim, Hyungjin

doi:10.1016/j.chaos.2022.111999

Cited by 11 publications

(8 citation statements)

References 64 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This implies that our research could offer valuable insights into the development of hardware systems to support clinical decisions. For instance, physical systems embodying SDU dynamics could be combined with neuromorphic devices [ [51] , [52] , [53] ], facilitating faster and more energy-efficient in-sensor intelligent diagnostics.…”

Section: Discussionmentioning

confidence: 99%

Robust QRS detection based on simulated degenerate optical parametric oscillator-assisted neural network

Liao,

Shi,

Sarker

et al. 2024

Heliyon

View full text Add to dashboard Cite

Section: Discussionmentioning

confidence: 99%

Robust QRS detection based on simulated degenerate optical parametric oscillator-assisted neural network

Liao,

Shi,

Sarker

et al. 2024

Heliyon

View full text Add to dashboard Cite

“…Besides, the stochastic current fluctuation by RTN, another intrinsic variation, can be also observed in memristive devices during the read operation due to the capture and emission of carriers in oxide traps. Especially, the RTN characteristics can be observed more frequently in HRS than LRS owing to longer tunneling distances caused by the rupture of conductive filaments [47,48]. This intrinsic variation can cause unstable read current characteristics when the inference operation is performed in a memristive neural network.…”

Section: Electrical Switching and Weight Transfer Characteristicsmentioning

confidence: 99%

Intrinsic variation effect in memristive neural network with weight quantization

Park¹,

Song²,

Youn³

et al. 2022

Nanotechnology

Self Cite

View full text Add to dashboard Cite

To analyze the effect of the intrinsic variations of the memristor device on the neuromorphic system, we fabricated 32×32 Al2O3/TiOx-based memristor crossbar array and implemented 3-bit multilevel conductance as weight quantization by utilizing the switching characteristics to minimize the performance degradation of the neural network. The tuning operation of 8 weight levels was confirmed with a tolerance of ±4 μA (±40 μS). The endurance and retention characteristics were also verified, and the random telegraph noise (RTN) characteristics were measured according to the weight range to evaluate the internal stochastic variation effect. Subsequently, a memristive neural network was constructed by off-chip training with differential memristor pairs for the modified National Institute of Standards and Technology (MNIST) handwritten dataset. The pre-trained weights were quantized, and the classification accuracy was evaluated by applying the intrinsic variations to each quantized weight. The intrinsic variations were applied using the measured weight inaccuracy given by the tuning tolerance, RTN characteristics, and the fault device yield. These results should be considered when the pre-trained weights are transferred to a memristive neural network during the off-chip training.

show abstract

“…Analogue-grade weight values can be realized with multilevel states between the HRS and LRS of the memristor device, and VMM operation can be conducted in parallel with applied voltage signals in a crossbar array in accordance with Ohm’s law and Kirchhoff’s current law. However, reliability issues persist in memristive devices, particularly concerning device-to-device (D2D) or cycle-to-cycle (C2C) variations arising due to the switching mechanism of memristive devices, which is predominantly based on the soft breakdown of dielectric layers. − Such variations can significantly degrade the performance of neural networks as inaccurate device states hinder precise VMM operations and lead to computation errors. Although a memristor array structure with active devices such as a transistor has been demonstrated to improve cell selectivity, − a passive crossbar array without active devices is advantageous for high-density integration of 4F 2 , thanks to the cross-point array structure. − With increasing cell resistance to mitigate IR drop caused by line resistances, the sneak path current issue in the passive crossbar array can be suppressed by designing bias schemes such as half-V and third-V schemes. , While the sneak path issue can be effectively suppressed by self-rectifying or monolithic integrable selectors, − it becomes negligible in a passive crossbar array as well during VMM operations because all the WLs and bitlines (BLs) are connected to known potential values, unlike stand-alone memory operations where most devices are on unselected WLs and BLs …”

Section: Introductionmentioning

confidence: 99%

Implementation of Convolutional Neural Networks in Memristor Crossbar Arrays with Binary Activation and Weight Quantization

Park,

Kim,

Song

et al. 2024

ACS Appl. Mater. Interfaces

Self Cite

View full text Add to dashboard Cite

We propose a hardware-friendly architecture of a convolutional neural network using a 32 × 32 memristor crossbar array having an overshoot suppression layer. The gradual switching characteristics in both set and reset operations enable the implementation of a 3-bit multilevel operation in a whole array that can be utilized as 16 kernels. Moreover, a binary activation function mapped to the read voltage and ground is introduced to evaluate the result of training with a boundary of 0.5 and its estimated gradient. Additionally, we adopt a fixed kernel method, where inputs are sequentially applied to a crossbar array with a differential memristor pair scheme, reducing unused cell waste. The binary activation has robust characteristics against device state variations, and a neuron circuit is experimentally demonstrated on a customized breadboard. Thanks to the analogue switching characteristics of the memristor device, the accurate vector−matrix multiplication (VMM) operations can be experimentally demonstrated by combining sequential inputs and the weights obtained through tuning operations in the crossbar array. In addition, the feature images extracted by VMM during the hardware inference operations on 100 test samples are classified, and the classification performance by off-chip training is compared with the software results. Finally, inference results depending on the tolerance are statistically verified through several tuning cycles.

show abstract

Effect of weight overlap region on neuromorphic system with memristive synaptic devices

Cited by 11 publications

References 64 publications

Robust QRS detection based on simulated degenerate optical parametric oscillator-assisted neural network

Robust QRS detection based on simulated degenerate optical parametric oscillator-assisted neural network

Intrinsic variation effect in memristive neural network with weight quantization

Implementation of Convolutional Neural Networks in Memristor Crossbar Arrays with Binary Activation and Weight Quantization

Contact Info

Product

Resources

About