基于深度学习技术的视神经系统研究

项目名称： 基于深度学习技术的视神经系统研究

项目编号： No.61771282

项目类型： 面上项目

立项/批准年度： 2018

项目学科： 无线电电子学、电信技术

项目作者： 宋健

作者单位： 清华大学

项目金额： 16万元

中文摘要： 计算神经学是随着脑神经科学进步与人工智能发展出现的跨信息科学和生命科学两领域的热点交叉学科。本项目吸收数学、物理等基础理论，使用信息科学理论与方法研究脑神经系统工作原理。该领域研究还尚缺乏系统性，在理论研究和实际应用中有一系列关键问题有待研究。.本项目主要从信息科学领域入手，针对具有可观测性的可见光信号对脑神经系统的影响进行研究。基于最有可能解开人脑/小鼠脑工作奥秘的深度学习方法，提出创新性的基于深度学习技术的视神经仿真系统架构，并结合参数的自适应均衡等技术，为实现高可靠性的视神经仿真和重构提供理论依据和技术支持。需要研究包括可见光调制、视神经信号采集、处理与建模、视觉处理脑区动态探测和仿真、神经电信号去噪、深度学习系统构建、重要参数适配、自适应调节技术、正反馈训练验证等一系列关键技术，以期取得突破。本项目团队具备交叉学科特性，研究具有良好的国际合作基础、初步研究成果，试验条件、经验和基础。

中文关键词： 深度学习；小鼠实验；视神经系统；计算神经学；生物信息系统仿真

英文摘要： With the progress of brain science and artificial intelligence, computational neuroscience is becoming an interdisciplinary subject between information and life science. This project includes mathematics, physics theories and information science theories that are used to study the working of brain nerve system. There is no systematic theoretical research and lack practical application. A series of key problems need to be researched.In this project, information sciences methods are used to explore the optical signal effects for the brain nerve system. Deep learning, the most likely answer to how human brain studies, is innovatively utilized to simulate the visual system. .To achieve the high reliability of the simulation on the optic nerve system, adaptive parameter equalization and reconstruction are integrated as the theoretical basis. In this project, visible light modulation, optic signal acquisition, processing and modeling, brain regions dynamic detection, processing and simulation, neural signal de noise, deep learning framework, major parameter adaptation, adaptive adjustment, positive feedback training will be thoroughly studied. The research team has established a very good international cooperation and has obtained preliminary research results for optical signal modulation as well as acquisition which lay down a solid basis for further studies.

英文关键词： deep learning;experiment in mice;optic nerve system;Computational neuroscience;System simulation of biological information