摘要
大脑作为人体的关键器官,其复杂程度非常高,被了解的也最少。为了研究与大脑相关的疾病和人类智能的机理,出现了新的学科——神经网络的仿真。神经网络仿真是和传统的神经网络算法完全不同的一个领域,属于信息学、生物学、计算机科学、仿真等多个领域相互交叉的学科。本文介绍了神经网络仿真的基本原理以及国外出现的开源工具和支撑项目,并对存在的仿真建模工具进行了全面的比较,以便于用户根据自身建模的需要选择合适的工具。文章对生物神经网络仿真的计算量进行了评估,并结合我国最新研制的天河1-A超级并行计算机讨论了并行生物神经网络仿真的前景。
As the most important part of human, brain is very complex. Because of the advantage to treat the disease of brain and understand the intelligence of humans, a new subject has emerged, i.e. the simulation of neural networks, which is a field that is very different to the traditional neural network algorithms. It is an interdisciplinary of informatics, biology, computer science and simulation. In this paper, the principle of the simulation of neural networks is introduced; open source toolboxes and relative projects are listed for the purpose to facilitate the user to choose suitable tools for his modeling work. The complexity is computed and the future of parallel simulation of neural networks is discussed based on the Tianhe1-A supercomputer.
出处
《计算机工程与科学》
CSCD
北大核心
2012年第1期137-142,共6页
Computer Engineering & Science
基金
国家自然科学基金资助项目(61003082
60873014)
关键词
仿真
神经网络
计算复杂性
simulation
neural networks
computing complexity