深層学習アルゴリズムを用いた皮質ネットワークの表現方法と計算処理の研究

使用深度学习算法研究皮层网络的表示方法和计算处理

• 批准号: 15F15351
• 负责人: ベヌッチ アンドレア
• 金额: $ 1.09万
• 依托单位: Institute of Physical and Chemical Research
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for JSPS Fellows
• 财政年份: 2015
• 资助国家: 日本
• 起止时间: 2015-10-09 至 2018-03-31
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-15F15351/
• 关键词: Mouse; Visual Cortex; Imaging; Decision Making; Neural Networks; Neural Computation; Dimensionality Reduction

Purpose of the research:The goal of this research proposal was to test the hypothesis that large-scale non-linear dynamical network models (i.e. recurrent neural networks - RNN) are an ideal mathematical framework to derive fundamental computational principles from the dynamics of large neuronal populations. We wanted to test this hypothesis in the context of perceptual learning (visual discrimination task). As a model system, we used the visual cortex of the mouse. This animal model has an unmatched set of experimental toolboxes to record and perturb the dynamics of large populations of cortical neurons.Achievements:The postdoctoral fellow awarded this prestigious fellowship, Dr. Nevrez Imamoglu had to leave the Laboratory for personal-related reasons. Given his Machine Learning background, Dr. Imamoglu spent a significant amount of time learning basic concepts in Systems Neuroscience, a novel field of research for him. Moreover, through continuous interactions with other Lab members, Dr. Imamoglu familiarized with the experimental procedures and the key characteristics of the neuronal data he was analyzing. Finally, he has provided important technical support to various aspects of other members’ projects, helping with software analysis and implementations. Unfortunately, regarding this fellowship proposal, Dr. Imamoglu had to leave the Laboratory before any progress could be made in terms of preliminary or publishable results.

研究目的：本研究提案的目标是测试大规模非线性动态网络模型（即递归神经网络- RNN）是从大型神经元群体的动态中导出基本计算原理的理想数学框架的假设。我们想在知觉学习（视觉辨别任务）的背景下测试这一假设。作为一个模型系统，我们使用了小鼠的视觉皮层。这个动物模型拥有一套无与伦比的实验工具箱，可以记录和扰动大量皮层神经元的动力学。成就：获得这一著名奖学金的博士后Nevrez Imamoglu博士因个人原因不得不离开实验室。鉴于他的机器学习背景，Imamoglu博士花了大量时间学习系统神经科学的基本概念，这对他来说是一个新的研究领域。此外，通过与其他实验室成员的持续互动，Imamoglu博士熟悉了实验程序和他正在分析的神经元数据的关键特征。最后，他为其他成员的项目的各个方面提供了重要的技术支持，帮助进行软件分析和实现。不幸的是，关于这项奖学金的建议，伊玛莫格鲁博士不得不离开实验室之前，可以在初步或可验证的结果方面取得任何进展。