Memristor devices have diverse physical models depending on their structure. In addition, the physical properties of memristors are described using complex differential equations. Therefore, it is necessary to integrate the various models of memristor into an unified physics-based model. In this paper, we propose a physics-informed neural network (PINN)-based compact memristor model. PINNs can solve complex differential equations intuitively and with ease. This methodology is used to conduct memristor physical analysis. The weight and bias extracted from the PINN are implemented in a Verilog-A circuit simulator to predict memristor device characteristics. The accuracy of the proposed model is verified using two memristor devices. The results show that PINNs can be used to extensively integrate memristor device models.
Download full-text PDF |
Source |
|---|---|
| http://www.ncbi.nlm.nih.gov/pmc/articles/PMC10892042 | PMC |
| http://dx.doi.org/10.3390/mi15020253 | DOI Listing |
Publication Analysis
Top Keywords
Similar Publications
Nonlinear memristive computational spectrometer.
Light Sci Appl
January 2025
State Key Laboratory of Infrared Physics, Shanghai Institute of Technical Physics, Chinese Academy of Sciences, 500 Yu-Tian Road, Shanghai, 200083, China.
In the domain of spectroscopy, miniaturization efforts often face significant challenges, particularly in achieving high spectral resolution and precise construction. Here, we introduce a computational spectrometer powered by a nonlinear photonic memristor with a WSe homojunction. This approach overcomes traditional limitations, such as constrained Fermi level tunability, persistent dark current, and limited photoresponse dimensionality through dynamic energy band modulation driven by palladium (Pd) ion migration.
View Article and Find Full Text PDFHighly Reliable BiOSe Dendritic Neuron Enabling Spatial-Temporal Signal Processing for Real-World Image Classification.
ACS Nano
January 2025
School of Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon 34141, Republic of Korea.
Artificial intelligence (AI) has made significant strides by imitating biological neurons and synapses through simplified models, yet incomplete neuron functionalities can limit performance and energy efficiency in handling complex tasks. Biological neurons process input signals nonlinearly, utilizing dendrites to process spatial-temporal information. This study demonstrates the compact artificial dendrite device employing memristors based on bismuth oxyselenide (BiOSe).
View Article and Find Full Text PDFOrientation-Selective Memory Switching in Quasi-1D NbSe Neuromorphic Device for Omnibearing Motion Detection.
Adv Mater
January 2025
School of Integrated Circuits, Harbin Institute of Technology (Shenzhen), Shenzhen, 518055, China.
Intelligent neuromorphic hardware holds considerable promise in addressing the growing demand for massive real-time data processing in edge computing. Resistive switching materials with intrinsic anisotropy and a compact design of non-volatile memory devices with the capability of handling spatiotemporally reconstructed data is crucial to perform sophisticated tasks in complex application scenarios. In this study, an anisotropic resistive switching cell with a planar configuration based on lithiated NbSe nanosheets is demonstrated.
View Article and Find Full Text PDFMemristors with Monolayer Graphene Electrodes Grown Directly on Sapphire Wafers.
ACS Appl Electron Mater
October 2024
School of Engineering and Materials Science, Queen Mary University of London, London E1 4NS, United Kingdom.
The development of the memristor has generated significant interest due to its non-volatility, simple structure, and low power consumption. Memristors based on graphene offer atomic monolayer thickness, flexibility, and uniformity and have attracted attention as a promising alternative to contemporary field-effect transistor (FET) technology in applications such as logic and memory devices, achieving higher integration density and lower power consumption. The use of graphene as electrodes in memristors could also increase robustness against degradation mechanisms, including oxygen vacancy diffusion to the electrode and unwanted metal ion diffusion.
View Article and Find Full Text PDFStochastic switching and analog-state programmable memristor and its utilization for homomorphic encryption hardware.
Nat Commun
July 2024
Department of Materials Science and Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Republic of Korea.
Article Synopsis
- Homomorphic encryption allows operations on encrypted data without needing to decrypt it, enhancing security during cloud-edge data communication.
- There is a rising demand for homomorphic encryption hardware (HE-HW) that is compact and energy-efficient for edge devices.
- The proposed Mo-MCM memristors utilize special properties to generate random numbers for encryption keys and store data, enabling key generation, encryption, and decryption all in one device through in-memory computing.
Want AI Summaries of new PubMed Abstracts delivered to your In-box?
Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!