A5004079546- Advanced Memory and Neural Computing
- Electromagnetic Compatibility and Noise Suppression
- Power Line Communications and Noise
- Semiconductor materials and devices
- Ferroelectric and Negative Capacitance Devices
- Fuel Cells and Related Materials
- Advanced Algorithms and Applications
- Electromagnetic Simulation and Numerical Methods
- Semiconductor Lasers and Optical Devices
- Advanced Neural Network Applications
- Numerical methods for differential equations
- Microwave Engineering and Waveguides
- Advanced Adaptive Filtering Techniques
- Advanced Power Amplifier Design
- Acoustic Wave Phenomena Research
- Probabilistic and Robust Engineering Design
Southern University of Science and Technology
2021-2024
Extracting behavioral models of RRAM devices is challenging due to their unique “memory” behaviors and rapid developments, for which well-established modeling frameworks systematic parameter extraction processes are not available. In this work, we propose a physics-informed recurrent neural network (PiRNN) methodology generate from practical measurement/simulation data. The proposed framework can faithfully capture the evolution internal state its impacts on output. A series modifications...
High-speed serial links are fundamental to energy-efficient and high-performance computing systems such as artificial intelligence, 5G mobile automotive, enabling low-latency high-bandwidth communication. Transmitters (TXs) within these key signal quality, while their modeling presents challenges due nonlinear behavior dynamic interactions with links. In this paper, we propose LiTformer: a Transformer-based model for high-speed link TXs, non-sequential encoder Transformer decoder incorporate...
Extracting behavioural models of RRAM devices has been challenging due to their unique "memory" behaviours, for which well-established modeling frameworks and systematic parameter extraction processes have not available. In this work, we propose a physics-informed recurrent neural network (PiRNN) methodology generate from practical measurement/simulation data. The proposed framework can faithfully capture the evolution internal state its impacts on output. A series modifications informed by...
Modeling of RRAM devices has been challenging due to the existence flux-dependent internal variables. Extra differential equations or capacitive circuits are often needed model evolution state variable and its impacts on device responses, rendering difficult develop slow evaluate. In this work, we propose a simple yet viable alternative build steady-state compact models using physics-informed neural networks that do not involve components. The central idea is utilize sequence currents...