I have been working on resistive memory device and novel computing concepts, such as analog computing and in-memory computing. My current research interests include:
- Analog matrix computing (AMC), including AMC circuits design with crosspoint RRAM arrays, fundamental theory about the circuits and the AMC concept, and their applications to machine learning, scientific computing, wireless communications and so on.
- Neural network models based on resistive switching dynamics, including dynamic characteristics and applications of RRAM circuits, and attractor network theory and algorithms.
- Novel resistive switching devices, including innovative design, optimization, and fabrication of devices and arrays, and design of accurate device programming mechanism.
My previous works include:
- Proposed an AMC circuit that can be used to solve a linear system Ax = b or an eigenvector equation Ax = lx in just one step. The circuit can be further used to solve differential equations, such as Fourier equation and Schrodinger equation, or to calculate PageRank scores, all in one step. Please refer to: Z. Sun, et al., PNAS (2019)；Z. Sun, et al., IEEE Trans. Electron Devices (2020)
- Revealed that the time complexity of the AMC circuits is validly O(1), outperforming its counterparts of both classical digital computing and quantum computing, which are O(polyN) and O(logN) respectively. Please refer to: Z. Sun, et al., IEEE Trans. Electron Devices (2020)；Z. Sun, et al., Adv. Intell. Syst. (2020)
- Proposed an AMC circuit that can be used to train traditional machine learning algorithms in one step, such as linear regression and logistic regression. The concept can be extended to polynomial regression, neural network and so on. Please refer to: Z. Sun, et al., Sci. Adv. (2020)
- Proposed the stateful neural network, which is the simplest neural network hardware architecture, and can be used to realized the most efficient stateful logic operation. Please refer to: Z. Sun, et al., Adv. Mater. (2018)
- Experimentally verified the dual-defect mechanism of metal-oxide resistive switching memory (or memristor), supporting the critical role of cation vacancy in the resistive switching mechanism, together with other contemporaneous results from some important labs. Please refer to: Z. Sun, et al., ACS Appl. Mater. Interfaces (2016)；A. Wedig, et al., Nat. Nanotechnol. (2016)；H. Jiang, et al., Sci. Rep. (2016)