类脑计算研究简述

A Brief Review of Brain-inspired Computing

类脑计算是借鉴脑信息处理方式,具备自主学习能力并擅于实时处理非结构化信息的新型计算智能,被认为可能是一条通向通用人工智能的途径。本文首先分析了类脑计算技术框架,并以此为基准,介绍了脑科学、脑图谱、各类脑仿真框架,并列举了常见的神经形态芯片和类脑计算系统;然后从模型、算法、框架、数据集方面系统地介绍了脉冲神经网络,以及当前类脑计算的应用情况;最后分析了类脑计算目前所存在的困难和挑战。

Brain-inspired computing is a new type of computing intelligence that learns from how brain processes information.It has the ability of self-learning and is good at processing unstructured information in real time.For that it’s thought to be a way towards artificial general intelligence.Firstly,we describe the technical framework of brain-inspired computing,and use it as a benchmark to introduce brain science,brain map and various types of brain simulation framework.Then we list common neuromorphic chips and brain-inspired computing system.Spiking neural networks are systematically elaborated from the aspects of models,algorithms,frameworks and datasets.The current applications of brain-inspired computing are also briefly discussed.Finally,the current difficulties and challenges of brain-inspired computing are analyzed.