脑科学视角下的高性能计算
发布时间：2019-04-03 13:58
【摘要】：高性能计算已经成为脑研究的必要手段,为脑研究开启了全新的研究视角和研究模式;反过来脑研究为高性能计算提供新思路、提出了新要求,催生着新型计算模式的出现.具体来说:(1)近年来脑观测技术性能不断提升,多模态观测普遍采用,全球范围内交叉联合研究也越来越成为常态,使得"大数据"和"云计算"成为脑研究的一种崭新研究视角和研究方法;(2)信息科学家提出了深度神经元网络、脉冲神经元网络等大规模学习算法,这些算法从神经科学中汲取了营养,显著提升了自身性能,实现了人脑智能的某些特征,具有大脑计算模式高效率、低功耗的特征,在多领域内取得了显著的成效.这些算法的研究和实现对高性能计算提出了越来越高的要求,给目前基于传统微处理器和计算架构的计算平台带来了几乎无法克服的困难;(3)信息科学家和神经科学家联合,开始在类脑微处理器和类脑计算架构等层面上开展新型类脑计算探索性研究,取得的研究进展显示了该领域未来充满希望.基于以上三点原因,高性能计算和脑科学研究正越来越紧密地交织在一起,彼此融入了对方的发展.文中以脑研究为切入点,分别从高性能计算对脑研究的技术支撑以及脑研究对高性能计算的启发和推动两个方面对脑科学视角下的高性能计算的现状和发展趋势进行了讨论.
[Abstract]:High-performance computing has become a necessary means of brain research, opening up a new research perspective and research model; in turn, brain research provides a new way of thinking for high-performance computing, puts forward new requirements, and spawns the emergence of new computing models. Specifically: (1) in recent years, the performance of brain observation technology has been continuously improved, multi-modal observation is widely used, and cross-joint research is becoming more and more normal all over the world. Make "big data" and "cloud computing" become a new research perspective and research method of brain research; (2) Information scientists have proposed large-scale learning algorithms such as deep neural network and pulse neural network. These algorithms draw nutrition from neuroscience, significantly improve their own performance, and realize some characteristics of human brain intelligence. With the characteristics of high efficiency and low power consumption of brain computing pattern, it has achieved remarkable results in many fields. The research and implementation of these algorithms put forward higher and higher requirements for high-performance computing, which brings almost insurmountable difficulties to the current computing platform based on traditional microprocessor and computing architecture. (3) Information scientists and neuroscientists began to carry out exploratory research on brain-like microprocessor and brain-like computing architecture. The research progress shows that the future of this field is full of hope. For the above three reasons, high-performance computing and brain science research are more and more closely intertwined and integrated into each other's development. Taking brain research as the starting point, this paper discusses the present situation and development trend of high performance computing in the perspective of brain science from two aspects: the technical support of high performance computing to brain research and the inspiration and promotion of brain research to high performance computing.
【作者单位】： 国防科学技术大学机电工程与自动化学院;国防科学技术大学国家高性能计算重点实验室;
【基金】：国家重大应用基础研究计划(2013CB329400) 国家自然科学基金(61473305,61375034)资助~~
【分类号】：R318;TP38
[Abstract]:High-performance computing has become a necessary means of brain research, opening up a new research perspective and research model; in turn, brain research provides a new way of thinking for high-performance computing, puts forward new requirements, and spawns the emergence of new computing models. Specifically: (1) in recent years, the performance of brain observation technology has been continuously improved, multi-modal observation is widely used, and cross-joint research is becoming more and more normal all over the world. Make "big data" and "cloud computing" become a new research perspective and research method of brain research; (2) Information scientists have proposed large-scale learning algorithms such as deep neural network and pulse neural network. These algorithms draw nutrition from neuroscience, significantly improve their own performance, and realize some characteristics of human brain intelligence. With the characteristics of high efficiency and low power consumption of brain computing pattern, it has achieved remarkable results in many fields. The research and implementation of these algorithms put forward higher and higher requirements for high-performance computing, which brings almost insurmountable difficulties to the current computing platform based on traditional microprocessor and computing architecture. (3) Information scientists and neuroscientists began to carry out exploratory research on brain-like microprocessor and brain-like computing architecture. The research progress shows that the future of this field is full of hope. For the above three reasons, high-performance computing and brain science research are more and more closely intertwined and integrated into each other's development. Taking brain research as the starting point, this paper discusses the present situation and development trend of high performance computing in the perspective of brain science from two aspects: the technical support of high performance computing to brain research and the inspiration and promotion of brain research to high performance computing.
【作者单位】： 国防科学技术大学机电工程与自动化学院;国防科学技术大学国家高性能计算重点实验室;
【基金】：国家重大应用基础研究计划(2013CB329400) 国家自然科学基金(61473305,61375034)资助~~
【分类号】：R318;TP38
【相似文献】
相关期刊论文 前10条
1 ;脑科学大事记[J];科学;1992年04期
2 杜久林,杨雄里;日本“脑科学时代”计划纲要简介[J];生命科学;1997年06期
3 冯聿峰;脑科学时代迎面走来[J];世界科技研究与发展;1999年06期
4 胡玉华;脑科学及其研究进展[J];北京教育学院学报;2000年01期
5 那杰;高飞;;日本“脑科学与教育”研究的进展[J];比较教育研究;2008年05期
6 李文管;;脑科学任务及人脑客体的角色地位[J];科学技术哲学研究;2010年02期
7 苏开源;日本推进脑科学十策[J];世界科技研究与发展;1999年06期
8 王剑;焦静;;好玩的脑科学[J];汽车消费报告;2010年01期
9 则鸣;毛文瑜;;新大陆:探秘脑科学[J];世界科学;2014年07期
10 陈惟昌,邱红霞,王自强;脑科学进展的跨世纪回顾与展望[J];世界科技研究与发展;2000年06期
相关重要报纸文章 前4条
1 时海t，
本文编号：2453270
本文链接：https://www.wllwen.com/yixuelunwen/swyx/2453270.html
