Physical Review Applied · 2022 · 11 citations · 34 references
EngineeringComputer EngineeringElectrical EngineeringMicroelectronicsMicrowave EngineeringSemiconductor DevicesDeep LearningDigital Image ProcessingImage ProcessingMemory DevicesMicrowave Semiconductor DevicesSpintronicsConvolutional Neural NetworksHandwritten Character RecognitionMicromagnetic ModelingMicromagnetic SimulationNeuromorphic DevicesNeuromorphic Hardware
Spintronic diodes (STDs) are emerging as a technology for the realization of high-performance microwave detectors. The key advantages of such devices are their high sensitivity, capability to work at low input power, and compactness. In this work, we show a possible use of STDs for neuromorphic computing to expand the realm of their functionalities for the implementation of analog multiplication, which is a key operation in convolutional neural networks (CNNs). In particular, we introduce the concept of degree of rectification (DOR) in injection-locked STDs. Micromagnetic simulations are used to design and identify the working range of the STDs for the implementation of the DOR. Previous experimental data confirm the applicability of the proposed solution, which is tested in image processing and in a CNN that recognizes handwritten digits.
34
Fully hardware-implemented memristor convolutional neural network
Peng Yao, Huaqiang Wu, Bin Gao et al. · Nature · 2020 · 2.1K citations
Engineering, Computer Engineering, Electrical Engineering +12
Parallel convolutional processing using an integrated photonic tensor core