A research team led by Academician Tiejun Cui and Associate Professor Qian Ma from the State Key Laboratory of Millimeter Waves at Southeast University has successfully developed a multilayer nonlinear diffractive neural network featuring a programmable, ultra-fast ReLU activation function. By integrating RF components into programmable metasurfaces to execute efficient ReLU-like operations, the system achieves nanosecond-level response latency, which is orders of magnitude faster than conventional optoelectronic conversion methods, while offering a low threshold and high reconfigurability. This work was published inNature Communications on November 24, 2025, under the title "Multilayer nonlinear diffraction neural networks with programmable and fast ReLU activation function," with Academician Tiejun Cui and Associate Professor Qian Ma as co-corresponding authors, and PhD student Yuming Ning and Associate Professor Qian Ma as co-first authors.
