作为下一代人工智能技术重大突破的热点方向，利用脉冲神经网络（SNN）构建类脑芯片进行神经形态计算的研究，面临着缺乏有效的学习方法、较差的预测性能和设计方法学的不完善等挑战。本项目针对目前该研究领域存在的问题和不足，开展新型脉冲神经网络的统计学习模型和硬件实现关键技术的研究。首先，避开大多数研究采取直接模仿生物表象的基于仿生的学习模型，本研究从探索生物神经网络内在的物理统计特点，建立基于生物似真性的统计学习模型；其次，结合集成电路技术，研究对统计学习和推理的各种电路技术实现，以确保统计学习模型得到真实而有效的实现；最后，在对学习模型和关键电路研究的基础上，建立协同模型算法和硬件实现进行脉冲神经形态计算芯片的跨层设计方法学。本研究涉及神经科学、人工智能和集成电路的跨学科研究，本质上是协同算法、结构和电路的跨层设计方法学的研究，为当今的神经形态计算研究探索开辟了一条新颖的、有特色的途径。

Emerging as a potential major breakthrough in the next-generation artificial intelligence (AI) technology, the research on neuromorphic computing using spiking neural network (SNN) to construct brain-like chips face major challenges, including lack of effective learning methods, poor prediction performance, and under-developed design methodology. In view of the existing problems and shortcomings in this research field, this research project carries out the research of an emerging SNN basd on novel statistical learning model and the key factors of hardware implementation. Firstly, instead of using conventional bio-mimcking based learning method that directly emulates biological phenomenon in most existing studies, this project explores the inherent physical and statistical characteristics of biological neural networks (SNN), and establishes a bio-plausible learning model for spiking neuromorphic computing hardware. Secondly, by utilizing advanced integrated circuit technology, we study various circuit designs and implementations for the novel statistical learning and reasoning, to ensure the loyal and efficient implementation of the statistical learning method. Finally, on the basis of the study of the new statistical learning model and the key implementation circuits, the research of joint algorithm and hardware design is established to study the cross-layer design methodology of the emerging spiking neuromorphic computing chip. In essence, this research is a cross-disciplinary study of cross-layer design methodology from algorithm, architecture, to circuit levels. This study opens up a novel and distinctive way for the research of neuromorphic computing computing hardware.