On June 20, a team led by Professor Miao Xiangshui and Professor Li Yi from the School of Integrated Circuit at Huazhong University of Science and Technology, in partnership with Professor Qian He and Professor Wu Huaqiang's team from Tsinghua University's School of Integrated Circuit, unveiled their groundbreaking research in memristive in-memory computing technology in the prestigious journal Science Advances. Their study, titled "Memristive Floating-Point Fourier Neural Operator Network for Efficient Scientific Modeling," introduces the world's first memristor-based floating-point precision neural network solver system. This system exemplifies high-precision, energy-efficient, and low-latency intelligent scientific computing, offering a potent solution to complex scientific modeling challenges, such as nonlinear differential equation models, and transcending the constraints of traditional numerical solvers.

The collaborative effort embodies a holistic optimization design spanning circuits, systems, and algorithms, thereby pioneering a novel approach for the efficient training and inference of neural network solvers.