High Performance Computing in the Viewpoint of Brain Research

Ya-Dong LIU,De-Wen HU
DOI: https://doi.org/10.11897/SP.J.1016.2017.02148
2017-01-01
Abstract:High performance computing has become a vital means of brain research,which opens up new research perspectives and patterns for it;In turn brain research gives rise to the emergence of a modern computing pattern by putting forward new requirements for high performance computing.Specifically:(1) in recent years,the performance of brain observation technology has steadily increased,the multimodal observation has been widely used,and the joint research worldwide has increasingly become the norm,thus "big data" and "cloud computing" become has been a new research perspective and methods of brain research;(2)large-scale learning algorithms such as large depth of deep neural network,spike neuron network,etc.were proposed by information scientists.These methods have learned much form neuroscience and improved their power significantly.The algorithms have high efficiency and lower power consumption characteristics of the brain computing model and implemented some of the characteristics of the human brain intelligence.And remarkable achievements have been made in many fields.The research and implementation of these algorithms have put forward higher requirements for high performance computing,which has brought almost insurmountable difficulties to the current computing platform based on the traditional microprocessor and computing architecture;(3) information scientists and neuroscientists have jointly carried out a new type of exploratory studies at the level of the microprocessor and brain computing architectures,and the progress has shown a promising future.Because of the above three reasons,high performance computing and brain science are closely interweaving together and being integrated into the development of each other.With the background of brain research,this paper discussed present status and development trend of both high performance computing technology support for brain research as well as the brain research's inspire and promotion to the high performance computing.
What problem does this paper attempt to address?