WO2022099557A1 - 一种类脑计算机操作系统的神经信息可视化方法 - Google Patents

一种类脑计算机操作系统的神经信息可视化方法 Download PDF

Info

Publication number
WO2022099557A1
WO2022099557A1 PCT/CN2020/128463 CN2020128463W WO2022099557A1 WO 2022099557 A1 WO2022099557 A1 WO 2022099557A1 CN 2020128463 W CN2020128463 W CN 2020128463W WO 2022099557 A1 WO2022099557 A1 WO 2022099557A1
Authority
WO
WIPO (PCT)
Prior art keywords
neuron
nodes
node
importance
brain
Prior art date
Application number
PCT/CN2020/128463
Other languages
English (en)
French (fr)
Inventor
李红
吕攀
李开心
杨国青
邓水光
潘纲
Original Assignee
浙江大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 浙江大学 filed Critical 浙江大学
Publication of WO2022099557A1 publication Critical patent/WO2022099557A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/451—Execution arrangements for user interfaces
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Definitions

  • the invention belongs to the technical field of novel computers, in particular to a neural information visualization method of a brain-like computer operating system.
  • Spiking Neural Networks (SNN-Spiking Neuron Networks) are often referred to as the third generation of artificial neural networks.
  • the first generation of neural networks was the perceptron, which was a simple neuron model and could only process binary data.
  • the second-generation neural network includes a relatively wide range, including the more widely used BP neural network. The spiking neural network is closer to the actual simulation of the connection relationship and behavior of brain neurons.
  • Netscope Take the model definition file as input to obtain the visual structure diagram of the neural network, which is a web-based visual neural network topology tool , which only supports UC Berkeley's caffe deep learning framework.
  • ConvNetDraw Use visualization commands as input to visualize the output neural network model structure, which is also a web-based tool. The model of the structural block is displayed, and the three-dimensional scale adjustment of the structural block can be carried out, which is very visual and intuitive.
  • Netron Take the model definition file and model weight file (can be default) as input to obtain the visual structure diagram of the neural network. It is also based on web pages and uses js and python to support ONNX, Keras, CoreML, TensorFlow, Mainstream deep learning frameworks such as caffe and MXNET.
  • Cao Lihong of Communication University of China disclosed a method for visualizing a large-scale neural network in 3D in the patent application with the publication number CN106372721A, showing the structure of the neural network in 3D; Qiu Chunfang of Shanghai Institute of Precision Measurement and Testing et al.
  • the purpose of the present invention is to provide a neural information visualization method of a brain-like computer operating system, so as to solve the problems of many connection description files, many neuron nodes, and too many connections between neuron nodes, which are difficult to distinguish.
  • a neural information visualization method of a brain-like computer operating system comprising the following steps:
  • connection description file of the spiking neural network, and globally encode the neuron nodes of each layer according to the connection description file.
  • Each neuron node obtains a global unique identifier, and each layer of neuron nodes is aggregated and represented by different colors.
  • the connection relationship of neuron nodes between levels is represented by connecting edges and distinguished by different colors;
  • all the neuron nodes in the same layer of the spiking neural network are collectively represented and represented by the same color, and the aggregated representations of the neuron nodes in the adjacent two layers are also in adjacent positions, which is convenient for the neuron nodes in the adjacent two layers. The connection relationship between them is represented.
  • each layer of neuron nodes is aggregated into geometric shapes for representation.
  • the geometric shape may be a smooth geometric shape without edges and corners such as a circle, an ellipse, or the like, or a geometric shape with edges and corners such as a matrix and a triangle.
  • connection relationships between the two layers of neuron nodes are represented by the same colored edges.
  • neuron nodes are colored differently than their corresponding connections.
  • the layout of neuron information is stored in a data structure
  • the data structure includes data units, connection relationship units, and layer units, wherein the data units are used to store the global unique identifiers of neuron nodes, Visualize the location coordinates and the level; the connection unit is used to store the visual color of two neuron nodes and the edges between the two neuron nodes; the layer unit is used to store the name of the level.
  • neuron nodes Due to the large number of neuron nodes, whether to display a neuron is selected according to the number of connections between the neuron node and other neurons and the weight of the connections. If the neuron nodes are sampled too much, the neuron nodes will be too dense after visualization, and the connection relationship between the neuron nodes cannot be seen; if the neuron nodes are sampled too little, the main information will be lost. The main structure of the spiking neural network cannot be observed. For reasonably sparse sampling of neurons, the importance of neuron nodes needs to be calculated and sorted. Preferably, the importance of each neuron node is calculated using the following formula:
  • In represents the absolute importance of the neuron before the regularization of the i-th neuron node node node i
  • InNum represents the in-degree of the neuron node
  • w 1 is the in-degree coefficient of the neuron node
  • OutNum represents the out-degree of the neuron node
  • w 2 is the out-degree coefficient of the neuron node
  • Weight represents the sum of the absolute value of the edge weight of the neuron node
  • w 3 is the weight coefficient of the neuron node
  • is the importance of the final neuron node represents the absolute importance of the neuron before the regularization of the jth neuron node node node j
  • n represents the total number of neuron nodes.
  • the number of connected edges connected to the current neuron node is taken as the in-degree of the current neuron node
  • the number of connected edges connected by the current neuron node is taken as the out-degree of the current neuron node.
  • an importance threshold is set, and neuron nodes whose importance is greater than the importance threshold are selected for visualization.
  • the more edges the neuron node is connected to the greater the absolute value of the edge weight will be preferentially reserved.
  • the beneficial effects of the present invention at least include:
  • the neural information visualization method of the brain-like computer operating system solves the problem of many connection description files by performing global coding on neurons, and screens and visualizes neurons according to the importance of neuron nodes.
  • the processing method solves the problem of difficulty in displaying multiple neuron nodes.
  • the connection relationship is represented in the form of edges and distinguished by different colors, which solves the problem that the connections between neuron nodes are difficult to distinguish.
  • the reasonable visualization of the neural information of the brain-like computer operating system is realized, which facilitates the understanding of the working process of the spiking neural network.
  • FIG. 1 is a flowchart of a method for visualizing neural information of a brain-like computer operating system provided by an embodiment of the present invention
  • Fig. 2 is the internal data structure of the connection description file of the spiking neural network provided by the embodiment of the present invention
  • 3 is an internal data structure of an intermediate json file for neural information visualization provided by an embodiment of the present invention.
  • Fig. 4 is the quadruple inside the data field of the json file provided by the embodiment of the present invention.
  • Fig. 5 is the triplet inside the json file links field provided by the embodiment of the present invention.
  • Fig. 6 is the neural information visualization diagram of the EEG simulation impulse neural network provided by the embodiment of the present invention.
  • FIG. 7 is a visualization diagram of neural information of a memory model spiking neural network provided by an embodiment of the present invention.
  • the embodiments of the present invention provide a method for visualizing neural information of a brain-like computer operating system. Description file for neural information visualization.
  • FIG. 1 is a flowchart of a method for visualizing neural information of a brain-like computer operating system provided by an embodiment of the present invention. As shown in Figure 1, the neural information visualization method includes the following steps:
  • Step 1 according to the connection relationship between the layers of the spiking neural network, read the connection description file of the spiking neural network in the order from the input layer to the output layer.
  • connection description file of the spiking neural network describes the connection relationship between the neurons and the neurons of the spiking neural network.
  • Figure 2 is the internal data structure of the connection description file of the spiking neural network. As shown in Figure 2, each line of the connection description file is a quadruple. The first field of the quadruple is the number in the starting neuron layer, that is, the number of the starting neuron in the starting layer.
  • the second field is the number in the target neuron layer, that is, the number of the target neuron in the target layer
  • the third field is the weight, that is, the weight of the neuron's processing input value, the weight is the inherent property of the neuron's edge
  • the fourth field is the spike delay, which is the delay of the spike signal from the starting neuron to the target neuron.
  • Step 2 according to the read spiking neural network connection description file, re-number the neuron nodes of each layer globally, convert the number in the layer into a global unique identifier, and determine the color of each layer of neurons to form the neuron's own attributes.
  • Step 3 Generate a triple (start neuron number, target neuron number, edge color) describing the node connection relationship according to the connection description file and the neuron node's own attributes. layer is determined.
  • Step 4 Layout the neuron nodes according to the connection relationship between neurons of the spiking neural network.
  • the neuron nodes of the same layer are gathered and placed in the same area, and the neurons of the adjacent layers are gathered and placed in adjacent positions.
  • Step 5 Select whether to display a neuron according to the number of connections between the neuron and other neurons and the weight of the connection.
  • the specific method is to calculate the importance of neurons according to the number of connections between neurons and other neurons and the weight of the connections. The neurons are sorted according to their importance, and neurons with greater importance will be preferentially reserved.
  • Step 6 export a json file describing the neuron layout and edge color, and use echarts to visualize the exported json file.
  • FIG. 3 is the data structure of the json file, which contains three fields: data unit data, connection relation unit links and layer unit layers.
  • the data field is an array, and each element inside is a quadruple.
  • the quadruple contains the key value name, position coordinates (x, y) and layer information of the neuron node, as shown in Figure 4.
  • the links field is also an array, each element inside is a triple, and the triple contains the key value source of the starting neuron node, the key value target of the target neuron, and the color lineStyle of the connection, as shown in the figure 5 shown.
  • the results of visualizing the neural information of the EEG simulated spiking neural network and the memory model spiking neural network are shown in Figures 6 and 7.
  • the figures clearly show the spiking neural network.
  • the structure of the network shows the neuron nodes of the spiking neural network and the connection relationship between each neuron node.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Neurology (AREA)
  • Computational Linguistics (AREA)
  • Microelectronics & Electronic Packaging (AREA)
  • Artificial Intelligence (AREA)
  • Human Computer Interaction (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • User Interface Of Digital Computer (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种类脑计算机操作系统的神经信息可视化方法，包括以下步骤：获得脉冲神经网络的连接描述文件，并根据连接描述文件对各层神经元节点进行全局编码，每个神经元节点获得一个全局唯一标识，每层神经元节点聚集表示且采用不同颜色区分，相邻层级之间神经元节点的连接关系用连边表示且采用不同颜色区分；根据神经元节点的连边数量和连边权重计算每个神经元节点的重要性，依据重要性大小选择神经元节点进行可视化。以解决连接描述文件多、神经元节点数量多、神经元节点间的连线多不易区分的问题。实现对类脑计算机操作系统的神经信息的合理可视化，方便对脉冲神经网络工作过程的理解。

Description

一种类脑计算机操作系统的神经信息可视化方法 技术领域
本发明属于新型计算机技术领域，具体涉及一种类脑计算机操作系统的神经信息可视化方法。
背景技术
深度学习自2006年产生之后就受到科研机构、工业界的高度关注，在图像和语音等多个领域取得了很大的发展，在多个领域全面超越传统算法。脉冲神经网络(SNN-Spiking Neuron Networks)经常被誉为第三代人工神经网络。第一代神经网络是感知器，它是一个简单的神经元模型并且只能处理二进制数据。第二代神经网络包括比较广泛，包括应用较多的BP神经网络。而脉冲神经网络更加接近实际的模拟了脑神经元的连接关系和行为。
目前，网络上公开的针对卷积神经网络结构可视化的工具较多，主要有：1)Netscope：以模型定义文件作为输入，得到神经网络的可视化结构图，是一个基于网页的可视化神经网络拓扑工具，仅仅支持加州大学伯克利分校的caffe深度学习框架。2)ConvNetDraw：使用可视化命令作为输入，可视化输出神经网络模型结构，也是一种基于网页的工具。以结构块的模型进行展现的，可以对结构块进行三个维度的比例调整，非常形象直观。3)Netron：以模型定义文件和模型权重文件(可以缺省)作为输入，获得神经网络的可视化结构图，同样是基于网页，并运用了js和python，能够支持ONNX、Keras、CoreML、TensorFlow、caffe、MXNET等主流 的深度学习框架。
中国传媒大学的曹力宏在公布号为CN106372721A的专利申请公开了一种以3D形式可视化大规模神经网络的方法，以3D的形式展示了神经网络的结构；上海精密计量测试研究所的邱春芳等人在公布号为CN107392085A的专利申请公开了一种可视化卷积神经网络的方法，能够可很好地展示出有助于了解卷积神经网络，用于探究卷积神经网络的优越性；北京计算机技术及应用研究所在公布号为CN110782031A的专利申请公开了一种多框架卷积神经网络模型结构可视化以及网络重建方法，可以直观的修改不同的层来实现网络的重建，同时能够更改神经网络属性并进行实时更新显示。
由于类脑计算机操作系统的脉冲神经网络结构的复杂性，在对类脑计算机操作系统的脉冲神经网络结构进行可视化时，会出现连接描述文件多、神经元节点数量多、神经元节点间的连线多不易区分的问题。因此，上述所有可视化方法均不适用于类脑计算机操作系统的脉冲神经网络结构的可视化。
发明内容
本发明的目的是提供一种类脑计算机操作系统的神经信息可视化方法，以解决连接描述文件多、神经元节点数量多、神经元节点间的连线多不易区分的问题。
为实现上述发明目的，本发明提供以下技术方案：
一种类脑计算机操作系统的神经信息可视化方法，包括以下步骤：
获得脉冲神经网络的连接描述文件，并根据连接描述文件对各层神经元节点进行全局编码，每个神经元节点获得一个全局唯一标识，每层神经 元节点聚集表示且采用不同颜色区分，相邻层级之间神经元节点的连接关系用连边表示且采用不同颜色区分；
根据神经元节点的连边数量和连边权重计算每个神经元节点的重要性，依据重要性大小选择神经元节点进行可视化。
优选地，处于脉冲神经网络同一层的所有神经元节点聚集表示，且采用相同颜色表示，相邻两层的神经元节点的聚集表示也处于相邻位置，便于相邻两层的神经元节点之间的连接关系表示。
优选地，每层神经元节点聚集成几何形状进行表示。所述几何形状可以为圆形、椭圆形等不具有棱角的光滑几何形状，还可以是矩阵、三角形等具有棱角的几何形状。
优选地，两层神经元节点之间的所有连接关系采用相同颜色连边表示。为了提升可视化清晰度，神经元节点的颜色与其对应的连边颜色不同。
优选地，在进行神经信息可视化时，神经元信息布局以一种数据结构存储，该数据结构包含数据单元、连接关系单元以及层单元，其中，数据单元用于存储神经元节点的全局唯一标识、可视化位置坐标以及所在层级；连接关系单元用于存储两个神经元节点和两个神经元节点之间连边的可视化颜色；层单元用于存储层级名称。
由于神经元节点的数量很多，根据神经元节点与其它神经元的连边数量和连边权重选择是否显示某个神经元。如果对神经元节点的采样过多，会造成可视化后神经元节点过于稠密，无法看出各个神经元节点之间的连接关系；如果对神经元节点的采样过少，会造成主要信息的丢失，不能观察到脉冲神经网络的主要结构。为了对神经元进行合理的稀疏化采样，需要计算神经元节点的重要性并进行排序。优选地，采用以下公式计算每个神经元节点的重要性：
Figure PCTCN2020128463-appb-000001
Figure PCTCN2020128463-appb-000002
其中，
Figure PCTCN2020128463-appb-000003
表示第i个神经元节点node i的正则化以前神经元的绝对重要性，InNum代表神经元节点的入度，w 1是神经元节点的入度系数，OutNum代表神经元节点的出度，w 2是神经元节点的出度系数，Weight代表神经元节点连边权重的绝对值之和，w 3是神经元节点的权重系数，
Figure PCTCN2020128463-appb-000004
是最终的神经元节点的重要性，
Figure PCTCN2020128463-appb-000005
表示第j个神经元节点node j的正则化以前神经元的绝对重要性，n表示神经元节点的总个数。其中，将接入当前神经元节点的连边个数作为当前神经元节点的入度，将当前神经元节点接出的连边个数作为当前神经元节点的出度。
优选地，依据重要性大小选择神经元节点进行可视化时，设置重要性阈值，筛选重要性大于该重要性阈值的神经元节点进行可视化。这样神经元节点连接到的边越多、边的权重绝对值越大会被优先保留。
与现有技术相比，本发明具有的有益效果至少包括：
本发明提供的类脑计算机操作系统的神经信息可视化方法，通过对神经元进行全局编码来解决连接描述文件多的问题，依据神经元节点的重要性对神经元进行筛选并可视化，采用这种稀疏化处理方法解决了神经元节点多显示困难的问题，同时将连接关系以连边形式表示且采用不同颜色区分，解决了神经元节点之间的连线多不易区分的问题。实现了对类脑计算机操作系统的神经信息的合理可视化，方便对对脉冲神经网络工作过程的 理解。
附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图做简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动前提下，还可以根据这些附图获得其他附图。
图1是本发明实施例提供的类脑计算机操作系统的神经信息可视化方法流程图；
图2是本发明实施例提供的脉冲神经网络的连接描述文件的内部数据结构；
图3是本发明实施例提供的用于神经信息可视化的中间json文件的内部数据结构；
图4是本发明实施例提供的json文件的data字段内部的四元组；
图5是本发明实施例提供的json文件links字段内部的三元组；
图6是本发明实施例提供的脑电模拟脉冲神经网络的神经信息可视化图；
图7是本发明实施例提供的记忆模型脉冲神经网络的神经信息可视化图。
具体实施方式
为使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例对本发明进行进一步的详细说明。应当理解，此处所描述的具体实施方式仅仅用以解释本发明，并不限定本发明的保护范围。
在对浙江大学达尔文2号神经拟态芯片的类脑计算机操作系统的脉冲神经网络进行神经信息可视化的过程中，遇到了连接描述文件多、神经元节点数量多、神经元节点间的连线多不易区分的问题。为了解决因为这些问题导致神经信息可视化困难的问题，本发明实施例提供了一种类脑计算机操作系统的神经信息可视化方法，基于达尔文2号神经拟态芯片的类脑计算机操作系统的脉冲神经网络模型连接描述文件进行神经信息可视化。
图1是本发明实施例提供的类脑计算机操作系统的神经信息可视化方法流程图。如图1所示，该神经信息可视化方法包括以下步骤：
步骤1，根据脉冲神经网络层与层的连接关系，按照从输入层到输出层的顺序读取脉冲神经网络的连接描述文件。
脉冲神经网络的连接描述文件描述了脉冲神经网络神经元和神经元之间的连接关系。脉冲神经网络的连接描述文件有多个，每个文件描述了两层神经元之间的连接关系。图2是脉冲神经网络的连接描述文件的内部数据结构。如图2所示，连接描述文件的每一行都是一个四元组，四元组的第一个字段是起始神经元层内编号，即起始神经元在起始层中的编号，第二个字段是目标神经元层内编号，即目标神经元在目标层中的编号，第三个字段为权重，即神经元的处理输入值的权重，该权重是神经元连边的固有属性，第四个字段是脉冲延时，即脉冲信号从起始神经元到目标神经元的延时。
步骤2，依据读取的脉冲神经网络连接描述文件，对各层神经元节点重新全局编号，将层内编号转化为全局唯一标识，同时确定每一层神经元的颜色，组成神经元自身属性。
步骤3，根据连接描述文件和神经元节点的自身属性生成描述节点连接关系的三元组(起始神经元编号，目标神经元编号，边的颜色)，边的 颜色根据边连接神经元所在的层确定。
步骤4，根据脉冲神经网络神经元之间的连接关系对神经元节点进行布局，同一层的神经元节点聚集放置在同一区域，相邻层的神经元聚集放置在相邻位置。
步骤5，根据神经元与其它神经元的连边数量和连边权重选择是否显示某个神经元。
由于原始的脉冲神经网络中的神经元节点和连线过多，如果全部显示，所有的节点和连线会挤在一起，不能展示出脉冲神经网络的关键结构，因此需要对脉冲神经网络进行稀疏化显示。如果对神经元的采样过多，会造成可视化后神经元过于稠密，无法看出各个神经元之间的连接关系；如果对神经元的采样过少，会造成主要信息的缺失，不能观察到脉冲神经网络的主要结构。具体的做法是根据神经元与其它神经元的连边数量和连边权重计算神经元的重要性，根据重要性对神经元进行排序，重要性越大的神经元会被优先保留。
神经元的重要性的公式如下：
Figure PCTCN2020128463-appb-000006
Figure PCTCN2020128463-appb-000007
在计算得到神经元的重要性以后，设置重要度阈值
Figure PCTCN2020128463-appb-000008
来控制显示哪些神经元，重要性大于阈值的神经元会被显示，小于阈值的神经元会被隐藏。
步骤6，导出描述神经元布局和边的颜色的json文件，利用echarts 将导出的json文件进行可视化。
实施例中，采用python代码生成json文件，采用echarts的graph类型对的json文件进行可视化。图3是json文件的数据结构，其中包含了三个字段：数据单元data，连接关系单元links和层单元layers。data字段是一个数组，内部的每个元素是一个四元组，四元组内包含了神经元节点的键值name、位置坐标(x，y)和所属层信息，如图4所示。links字段同样是一个数组，内部的每个元素是一个三元组，三元组内部包含了起始神经元节点的键值source，目标神经元的键值target，连线的颜色lineStyle，如图5所示。
利用上述类脑计算机操作系统的神经信息可视化方法方法，对脑电模拟脉冲神经网络和记忆模型脉冲神经网络神经信息可视化的结果如图6和图7所示，图中清晰地展示出了脉冲神经网络的结构，即展示出脉冲神经网络的神经元节点和各个神经元节点之间的连接关系。
以上所述的具体实施方式对本发明的技术方案和有益效果进行了详细说明，应理解的是以上所述仅为本发明的最优选实施例，并不用于限制本发明，凡在本发明的原则范围内所做的任何修改、补充和等同替换等，均应包含在本发明的保护范围之内。

Claims (8)

  1. 一种类脑计算机操作系统的神经信息可视化方法，其特征在于，包括以下步骤：
    获得脉冲神经网络的连接描述文件，并根据连接描述文件对各层神经元节点进行全局编码，每个神经元节点获得一个全局唯一标识，每层神经元节点聚集表示且采用不同颜色区分，相邻层级之间神经元节点的连接关系用连边表示且采用不同颜色区分；
    根据神经元节点的连边数量和连边权重计算每个神经元节点的重要性，依据重要性大小选择神经元节点进行可视化。
  2. 如权利要求1所述的类脑计算机操作系统的神经信息可视化方法，其特征在于，处于脉冲神经网络同一层的所有神经元节点聚集表示，且采用相同颜色表示，相邻两层的神经元节点的聚集表示也处于相邻位置，便于相邻两层的神经元节点之间的连接关系表示。
  3. 如权利要求1或2所述的类脑计算机操作系统的神经信息可视化方法，其特征在于，每层神经元节点聚集成几何形状进行表示。
  4. 如权利要求1所述的类脑计算机操作系统的神经信息可视化方法，其特征在于，两层神经元节点之间的所有连接关系采用相同颜色连边表示。
  5. 如权利要求1所述的类脑计算机操作系统的神经信息可视化方法，其特征在于，神经元节点的颜色与其对应的连边颜色不同。
  6. 如权利要求1所述的类脑计算机操作系统的神经信息可视化方法，其特征在于，在进行神经信息可视化时，神经元信息布局以一种数据结构存储，该数据结构包含数据单元、连接关系单元以及层单元，其中，数据单元用于存储神经元节点的全局唯一标识、可视化位置坐标以及所在层级； 连接关系单元用于存储两个神经元节点和两个神经元节点之间连边的可视化颜色；层单元用于存储层级名称。
  7. 如权利要求1所述的类脑计算机操作系统的神经信息可视化方法，其特征在于，采用以下公式计算每个神经元节点的重要性：
    Figure PCTCN2020128463-appb-100001
    Figure PCTCN2020128463-appb-100002
    其中，
    Figure PCTCN2020128463-appb-100003
    表示第i个神经元节点node i的正则化以前神经元的绝对重要性，InNum代表神经元节点的入度，w 1是神经元节点的入度系数，OutNum代表神经元节点的出度，w 2是神经元节点的出度系数，Weight代表神经元节点连边权重的绝对值之和，w 3是神经元节点的权重系数，
    Figure PCTCN2020128463-appb-100004
    是最终的神经元节点的重要性，
    Figure PCTCN2020128463-appb-100005
    表示第j个神经元节点node j的正则化以前神经元的绝对重要性，n表示神经元节点的总个数。
  8. 如权利要求1所述的类脑计算机操作系统的神经信息可视化方法，其特征在于，依据重要性大小选择神经元节点进行可视化时，设置重要性阈值，筛选重要性大于该重要性阈值的神经元节点进行可视化。
PCT/CN2020/128463 2020-11-11 2020-11-12 一种类脑计算机操作系统的神经信息可视化方法 WO2022099557A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011255463.4 2020-11-11
CN202011255463.4A CN112270406B (zh) 2020-11-11 2020-11-11 一种类脑计算机操作系统的神经信息可视化方法

Publications (1)

Publication Number Publication Date
WO2022099557A1 true WO2022099557A1 (zh) 2022-05-19

Family

ID=74340656

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/128463 WO2022099557A1 (zh) 2020-11-11 2020-11-12 一种类脑计算机操作系统的神经信息可视化方法

Country Status (2)

Country Link
CN (1) CN112270406B (zh)
WO (1) WO2022099557A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115879544A (zh) * 2023-02-28 2023-03-31 中国电子科技南湖研究院 一种针对分布式类脑仿真的神经元编码方法及系统

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112819160B (zh) * 2021-02-24 2023-10-31 文远鄂行(湖北)出行科技有限公司 一种神经网络模型的可视化方法、装置、设备及存储介质
CN112561043B (zh) * 2021-03-01 2021-06-29 浙江大学 一种类脑计算机操作系统的神经模型拆分方法
CN112561042B (zh) * 2021-03-01 2021-06-29 浙江大学 一种类脑计算机操作系统的神经模型映射方法
CN113297314B (zh) * 2021-07-28 2021-11-02 深圳市永达电子信息股份有限公司 一种数据可视化方法及装置、存储介质
CN113791913B (zh) * 2021-11-16 2022-02-11 浙江大学 一种类脑计算机操作系统的任务迁移方法
CN114399033B (zh) * 2022-03-25 2022-07-19 浙江大学 基于神经元指令编码的类脑计算系统和计算方法
CN114611686B (zh) * 2022-05-12 2022-08-30 之江实验室 基于可编程神经拟态核的突触延时实现系统及方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150106306A1 (en) * 2013-10-16 2015-04-16 University Of Tennessee Research Foundation Method and apparatus for constructing a neuroscience-inspired artificial neural network with visualization of neural pathways
CN107704925A (zh) * 2017-10-16 2018-02-16 清华大学 深度神经网络训练过程的可视分析系统及方法
CN110782031A (zh) * 2019-09-27 2020-02-11 北京计算机技术及应用研究所 多框架卷积神经网络模型结构可视化以及网络重建方法

Family Cites Families (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20150085007A (ko) * 2012-11-13 2015-07-22 엘마인다 리미티드 시공간 분할을 이용한 신경 생리학적 데이터 분석
CA2979579C (en) * 2015-03-20 2020-02-18 Fraunhofer-Gesellschaft Zur Foerderung Der Angewandten Forschung E.V. Relevance score assignment for artificial neural networks
US10318584B2 (en) * 2015-05-26 2019-06-11 Ayasdi, Inc. Outcome analysis for graph generation
CN105788608B (zh) * 2016-03-03 2019-03-26 渤海大学 基于神经网络的汉语声韵母可视化方法
US11086471B2 (en) * 2016-06-06 2021-08-10 Salesforce.Com, Inc. Visualizing neural networks
CN106372721B (zh) * 2016-08-29 2018-08-21 中国传媒大学 大规模神经网络的3d可视化方法
US11544539B2 (en) * 2016-09-29 2023-01-03 Tsinghua University Hardware neural network conversion method, computing device, compiling method and neural network software and hardware collaboration system
CN106650922B (zh) * 2016-09-29 2019-05-03 清华大学 硬件神经网络转换方法、计算装置、软硬件协作系统
US20190005384A1 (en) * 2017-06-29 2019-01-03 General Electric Company Topology aware graph neural nets
US10366322B2 (en) * 2017-10-06 2019-07-30 DeepCube LTD. System and method for compact and efficient sparse neural networks
CN110322010B (zh) * 2019-07-02 2021-06-25 深圳忆海原识科技有限公司 用于类脑智能与认知计算的脉冲神经网络运算系统及方法
CN110781933B (zh) * 2019-10-14 2022-08-05 杭州电子科技大学 一种用于理解图卷积神经网络的可视分析方法
CN111082949B (zh) * 2019-10-29 2022-01-28 广东工业大学 一种类脑计算机中脉冲数据包高效传输方法
CN110909869B (zh) * 2019-11-21 2022-08-23 浙江大学 一种基于脉冲神经网络的类脑计算芯片
CN111291243B (zh) * 2019-12-30 2022-07-12 浙江大学 一种人物事件的时空信息不确定性的可视化推理方法
CN111797333B (zh) * 2020-06-04 2021-04-20 南京擎盾信息科技有限公司 舆情传播任务展示的方法及装置

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150106306A1 (en) * 2013-10-16 2015-04-16 University Of Tennessee Research Foundation Method and apparatus for constructing a neuroscience-inspired artificial neural network with visualization of neural pathways
CN107704925A (zh) * 2017-10-16 2018-02-16 清华大学 深度神经网络训练过程的可视分析系统及方法
CN110782031A (zh) * 2019-09-27 2020-02-11 北京计算机技术及应用研究所 多框架卷积神经网络模型结构可视化以及网络重建方法

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
ELBEZ HAMMOUDA, BENHAOUA KAMEL, DEVIENNE PHILIPPE, BOULET PIERRE: "Visualization Techniques in SNN Simulators", 3RD INTERNATIONAL CONFERENCE ON MULTIMEDIA INFORMATION PROCESSING, 31 October 2018 (2018-10-31), pages 1 - 6, XP055931232 *
KASABOV NIKOLA K.: "NeuCube: A spiking neural network architecture for mapping, learning and understanding of spatio-temporal brain data", NEURAL NETWORKS, vol. 52, 1 April 2014 (2014-04-01), GB , pages 62 - 76, XP055931225, ISSN: 0893-6080, DOI: 10.1016/j.neunet.2014.01.006 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115879544A (zh) * 2023-02-28 2023-03-31 中国电子科技南湖研究院 一种针对分布式类脑仿真的神经元编码方法及系统
CN115879544B (zh) * 2023-02-28 2023-06-16 中国电子科技南湖研究院 一种针对分布式类脑仿真的神经元编码方法及系统

Also Published As

Publication number Publication date
CN112270406B (zh) 2023-05-23
CN112270406A (zh) 2021-01-26

Similar Documents

Publication Publication Date Title
WO2022099557A1 (zh) 一种类脑计算机操作系统的神经信息可视化方法
Zhou et al. Edge bundling in information visualization
CN106778894A (zh) 一种学术异构信息网络中作者合作关系预测的方法
US20100277481A1 (en) Method and apparatus of animation planning for a dynamic graph
CN106295799A (zh) 一种深度学习多层神经网络的实现方法
CN110414718A (zh) 一种深度学习下的配电网可靠性指标优化方法
CN110009108A (zh) 一种全新的量子超限学习机
CN115222792A (zh) 一种铁路桥梁数字孪生建模方法
CN110782031A (zh) 多框架卷积神经网络模型结构可视化以及网络重建方法
CN106789149A (zh) 采用改进型自组织特征神经网络聚类算法的入侵检测方法
Li et al. Fault identification in power network based on deep reinforcement learning
CN115659985A (zh) 电力知识图谱实体对齐方法、装置和计算机设备
CN104778308B (zh) 飞机结构型材的识别方法和装置
Shi et al. A fault location method for distribution system based on one-dimensional convolutional neural network
Wu et al. Active 3-D shape cosegmentation with graph convolutional networks
CN110956010B (zh) 基于梯度提升树的大规模新能源接入电网稳定性辨识方法
CN106650916A (zh) 一种基于蚁群优化的网格分割方法
Su et al. A multi-level visualization method for IT system structure
Peng et al. Dmnevis: a novel visual approach to explore evolution of dynamic multivariate network
Liu et al. RankAxis: Towards a systematic combination of projection and ranking in multi-attribute data exploration
CN109635848A (zh) 一种基于特征聚类的深度学习方法
Xi et al. Adaptive-Correlation-aware Unsupervised Deep Learning for Anomaly Detection in Cyber-physical Systems
Jiang et al. Predictive Methods for the Direction of Convergence in Emerging Industrial Technologies
Mao et al. Model Analysis of Applying Computer Monitoring to College Students’ Mental Health
Cho et al. Visualization of convolutional neural networks for time series input data

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20961114

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20961114

Country of ref document: EP

Kind code of ref document: A1