An Improved Cellular Nonlinear Network Architecture for Binary and Grayscale Image Processing

Müller, Jens; Wittig, Robert; Müller, J. Steffen; Tetzlaff, Ronald

doi:10.1109/tcsii.2016.2621773

Cited by 11 publications

(4 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The maximum size of dimensions of PE is dependent on the width of the buffer length register (WBLR) of DMA Controller, WBLR can be set from 8 bits to 23 bits. Therefore, if the data width of DMA Controller is 32 bits, the maximum size of dimensions is 2 21 -1, which is 2,097,151. As is shown in Table 2, the time consumption is increased with increase in dimensions.…”

Section: Changeable-dimension Matrix Computing Testmentioning

confidence: 99%

An Accelerator Architecture of Changeable-Dimension Matrix Computing Method for SVM

Liu

et al. 2019

Electronics

View full text Add to dashboard Cite

Matrix multiplication is a critical time-consuming processing step in many machine learning applications. Due to the diversity of practical applications, the matrix dimensions are generally not fixed. However, most matrix calculation methods, based on field programmable gate array (FPGA) currently use fixed matrix dimensions, which limit the flexibility of machine learning algorithms in a FPGA. The bottleneck lies in the limited FPGA resources. Therefore, this paper proposes an accelerator architecture for matrix computing method with changeable dimensions. Multi-matrix synchronous calculation concept allows matrix data to be processed continuously, which improves the parallel computing characteristics of FPGA and optimizes the computational efficiency. This paper tests matrix multiplication using support vector machine (SVM) algorithm to verify the performance of proposed architecture on the ZYNQ platform. The experimental results show that, compared to the software processing method, the proposed architecture increases the performance by 21.18 times with 9947 dimensions. The dimension is changeable with a maximum value of 2,097,151, without changing hardware design. This method is also applicable to matrix multiplication processing with other machine learning algorithms.

show abstract

Section: Changeable-dimension Matrix Computing Testmentioning

confidence: 99%

An Accelerator Architecture of Changeable-Dimension Matrix Computing Method for SVM

Liu

et al. 2019

Electronics

View full text Add to dashboard Cite

show abstract

“…The modulation of resistive memory devices by light has numerous applications including light-switchable memory elements in optical communications and photonic integrated circuits, fast in-memory computer vision systems, , and light tunable artificial synapses in neuromorphic computing. − Recently, we have pioneered several optical memristor types based on polymers mixed with metal nanoparticles, ZnO nanorods (NRs), and graphene oxide . The use of hybrid materials in memristor devices is generally much less studied than their solid-state counterparts.…”

Section: Introductionmentioning

confidence: 99%

Polymer/TiO₂ Nanorod Nanocomposite Optical Memristor Device

et al. 2021

View full text Add to dashboard Cite

Modulation of resistive switching memory by light opens the route to new optoelectronic devices that can be controlled both optically and electronically. Applications include integrated circuits with memory elements switchable by light and neuromorphic computing with optically reconfigurable and tunable synaptic circuits. We report on a unique nanocomposite resistive switching material and device made from a low concentration (∼0.1% by mass) of titanium dioxide nanorods (TiO 2 -NRs) embedded within the azobenzene polymer, poly(disperse red 1 acrylate, PDR1A). The device exhibits both reversible electronic memristor switching and reversible polarization-dependent optical switching. Optical irradiation by circularly polarized light causes a trans−cis photochemical isomerization that modifies the conformation and orientation of the photoactive azo-unit within the polymer. The resulting expansion of the composite (PDR1A/TiO 2 -NR) polymer film modifies the conduction pathway, facilitated by the presence of the TiO 2 -NRs, as a semiconductor material, through the (PDR1A/TiO 2 -NR) polymer film, which provides a sensitive means to control resistive switching in the device. The effect is reversible by changing the polarization state of the incident light. A charge-flux memristor model successfully reproduces the current−voltage hysteresis loops and threshold switching properties of the device, as well as the effect of the illumination on the electrical characteristics.

show abstract

“…However there are significant advantages if devices can as well be switched by optical means since optical signalling supports higher bandwidth communication, faster transmission speeds and the ability to decouple signals from electronic noise. A suite of potential new applications include; switchable memory elements in optical communications and photonic integrated circuits, optically tunable synaptic elements and new types of light sensors that can be used in cellular neural network cameras [9][10][11]. The latter having the important advantage that some in-memory processing of the image (e.g.…”

Section: Introductionmentioning

confidence: 99%

Wavelength dependent light tunable resistive switching graphene oxide nonvolatile memory devices

Jaafar

Kemp

2019

Carbon

View full text Add to dashboard Cite

This paper reports on the first optically tunable graphene oxide memristor device. Modulation of resistive switching memory by light opens the route to new optoelectronic devices that can be switched optically and read electronically. Applications include integrated circuits with memory elements switchable by light and optically reconfigurable and tunable synaptic circuits for neuromorphic computing and braininspired, artificial intelligence systems. In this report, planar and vertical structured optical resistive switching memristors based on graphene oxide are reported. The device is switchable by either optical or electronic means, or by a combination of both. In addition the devices exhibit a unique wavelength dependence that produces reversible and irreversible properties depending on whether the irradiation is long or short wavelength light, respectively. For long wavelength light, the reversible photoconductance effect permits short-term dynamic modulation of the resistive switching properties of the light, which has application as short-term memory in neuromorphic computing. In contrast, short wavelength length induces both a reversible photoconductance effect and irreversible changes due to reduction of the graphene oxide. This has important application in the fabrication of neural networks with factory defined weights, enabling the fast replication of artificial intelligent chips with pre-trained information.

show abstract

An Improved Cellular Nonlinear Network Architecture for Binary and Grayscale Image Processing

Cited by 11 publications

References 17 publications

An Accelerator Architecture of Changeable-Dimension Matrix Computing Method for SVM

An Accelerator Architecture of Changeable-Dimension Matrix Computing Method for SVM

Polymer/TiO₂ Nanorod Nanocomposite Optical Memristor Device

Wavelength dependent light tunable resistive switching graphene oxide nonvolatile memory devices

Contact Info

Product

Resources

About

An Improved Cellular Nonlinear Network Architecture for Binary and Grayscale Image Processing

Cited by 11 publications

References 17 publications

An Accelerator Architecture of Changeable-Dimension Matrix Computing Method for SVM

An Accelerator Architecture of Changeable-Dimension Matrix Computing Method for SVM

Polymer/TiO2 Nanorod Nanocomposite Optical Memristor Device

Wavelength dependent light tunable resistive switching graphene oxide nonvolatile memory devices

Contact Info

Product

Resources

About

Polymer/TiO₂ Nanorod Nanocomposite Optical Memristor Device