Press "Enter" to skip to content

PyVis|神经网络数据集的可视化

本站内容均来自兴趣收集,如不慎侵害的您的相关权益,请留言告知,我们将尽快删除.谢谢.

文章目录

 

pyvis 0.2.1

 

应用实例:

 

我自己的仓库: https://github.com/shandianchengzi/tree_sitter_DataExtractor 。根目录的 test1.html 就是测试结果, show_data.py 是本篇博客主要涉及的代码。

 

可视化结果:

 

PyVis的应用:

 

零、官方教程

 

https://pyvis.readthedocs.io/en/latest/tutorial.html

 

一、初始化画布 Network

 

默认设置的初始化方式(长宽500px、无向图):

 

>>> from pyvis.network import Network
>>> net = Network()

 

其中 Network() 里可以加自定义参数,如长宽1000px、有向图:

 

net = Network('1000px', '1000px', directed=True)

 

可以修改的参数有:

 

参数名默认值说明
height500px画布高度
width500px画布宽度
directedFalse是否是有向图
notebookFalseTrue if using jupyter notebook.
bgcolor#ffffff画布背景颜色
font_colorFalseThe color of the node labels text
layoutNoneUse hierarchical layout if this is set

 

授人以鱼不如授人以渔,这些参数、默认值以及说明全部都可以直接在pyvis的源代码中看到: https://github.com/WestHealth/pyvis/blob/master/pyvis/network.py

 

二、添加结点

 

添加单个结点 add_node

 

>>> net.add_node(1, label="Node 1") # node id = 1 and label = Node 1
>>> net.add_node(2) # node id and label = 2

 

支持的参数:

 

n_id: 唯一标识(str or int)
label: 结点标签。
以上必须,以下可选。
· borderWidth: The width of the border of the node.
· borderWidthSelected: The width of the border of the node when it is selected. When undefined, the borderWidth * 2 is used.
· brokenImage: 图片结点(shape=image或circularImage)加载不出来时,显示的URL或文字说明。
· group: 如果未定义,则节点将属于已经定义的组。该组的样式信息将应用于此节点。特定于节点的样式会覆盖组样式。
· hidden: When true, the node will not be shown. It will still be part of the physics simulation though! 
· image: When the shape is set to image or circularImage, this option should be the URL to an image. If the image cannot be found, the brokenImage option can be used.
· labelHighlightBold: Determines whether or not the label becomes bold when the node is selected.
· level: When using the hierarchical layout, the level determines where the node is going to be positioned.
· mass: The barnesHut physics model (which is enabled by default) is based on an inverted gravity model. By increasing the mass of a node, you increase it's repulsion. Values lower than 1 are not recommended.
· physics: When false, the node is not part of the physics simulation. It will not move except for from manual dragging.
· shape: The shape defines what the node looks like. There are two types of nodes. One type has the label inside of it and the other type has the label underneath it. The types with the label inside of it are: ellipse, circle, database, box, text. The ones with the label outside of it are: image, circularImage, diamond, dot, star, triangle, triangleDown, square and icon.
· size: The size is used to determine the size of node shapes that do not have the label inside of them. These shapes are: image, circularImage, diamond, dot, star, triangle, triangleDown, square and icon.
· title: 鼠标移到结点上方时,会显示的内容。hover-text。
· value: When a value is set, the nodes will be scaled using the options in the scaling object defined above.
· x: This gives a node an initial x position. When using the hierarchical layout, either the x or y position is set by the layout engine depending on the type of view. The other value remains untouched. When using stabilization, the stabilized position may be different from the initial one. To lock the node to that position use the physics or fixed options.
· y: This gives a node an initial y position. When using the hierarchical layout,either the x or y position is set by the layout engine depending on the type of view. The other value remains untouched. When using stabilization, the stabilized position may be different from the initial one. To lock the node to that position use the physics or fixed options.

 

更详细的参数赋值说明直接看源代码: https://github.com/WestHealth/pyvis/blob/master/pyvis/network.py

 

添加一系列结点 add_nodes

 

可选参数变少了,只支持参数 ["size", "value", "title", "x", "y", "label", "color", "shape"]

 

应用样例如下:

 

net.add_nodes(range(node_nums),
                 title=[str(x) for x in range(node_nums)],
                 label=[name[str(x)] for x in range(node_nums)],
                 color=color
                 )

 

三、添加边

 

添加一条边 add_edge

 

>>> net.add_node(0, label='a')
>>> net.add_node(1, label='b')
>>> net.add_edge(0, 1)

 

支持的参数:

 

· arrowStrikethrough: When false, the edge stops at the arrow. This can be useful if you have thick lines and you want the arrow to end in a point. Middle arrows are not affected by this.
· from: Edges are between two nodes, one to and one from. This is where you define the from node. You have to supply the corresponding node ID. This naturally only applies to individual edges.
· hidden: When true, the edge is not drawn. It is part still part of the physics simulation however!
· physics: When true, the edge is part of the physics simulation. When false, it will not act as a spring.
· title: The title is shown in a pop-up when the mouse moves over the edge.
· to: Edges are between two nodes, one to and one from. This is where you define the to node. You have to supply the corresponding node ID. This naturally only applies to individual edges.
· value: When a value is set, the edges' width will be scaled using the options in the scaling object defined above.
· width: The width of the edge. If value is set, this is not used.

 

添加多条边 add_edges

 

>>> edges = [[0,1],[1,2,2],[2,3]]
>>> net.add_edges(edges)

 

只支持width参数,它的实现是这样的,相当鸡肋,感觉可能是作者懒得写了:

 

def add_edges(self, edges):
    for edge in edges:
        # if incoming tuple contains a weight
        if len(edge) == 3:
            self.add_edge(edge[0], edge[1], width=edge[2])
        else:
            self.add_edge(edge[0], edge[1])

 

四、为结点添加漂亮的颜色(从而表明结点顺序)

 

color = ['#00ffff']  # 为第一个结点赋予特殊颜色
for x in range(node_nums - 1):
    current_of_color = int(x / node_nums * 255 * 2)
    if current_of_color > 255:
        color.append('#%02x%02x%02x' % (255, current_of_color - 255, current_of_color - 255))
    else:
        color.append('#%02x%02x%02x' % (current_of_color, 0, 0))

 

参考: 《如何实现从一种颜色的渐变效果(如从深红-红-浅红),RGB()不好用,但又没有SHL()这个函数。》-CSDN-帖子

 

完整代码(仅供参考):

 

from pyvis.network import Network
import gzip
import json
def load_graph(path):
    with gzip.open(path, 'r') as f:
        content = f.read().decode('utf-8')
        context_graph = []
        items = content.split('
')
        for item in items:
            try:
                context_graph.append(json.loads(item))
            except:
                print(item)
        return context_graph
# input:path:数据集路径;index:数据集条目索引;labels:需要显示的额外边的名称
# 将数据集的某条数据可视化
def show_graph(path, index, *labels):
    # 1. 载入数据
    graph = load_graph(path)
    one_data = graph[index]
    child_edges = one_data['ContextGraph']['Edges']['Child']
    node_nums = len(one_data['ContextGraph']['NodeLabels'])
    # 2. 为结点按顺序选取不同的颜色(从深色冷色调到浅色暖色调,并为首个结点赋予特殊颜色,这种方法在结点超过255*2后相邻结点颜色可能重复)
    color = ['#00ffff']  # 为第一个结点赋予特殊颜色
    for x in range(node_nums - 1):
        current_of_color = int(x / node_nums * 255 * 2)
        if current_of_color > 255:
            color.append('#%02x%02x%02x' % (255, current_of_color - 255, current_of_color - 255))
        else:
            color.append('#%02x%02x%02x' % (current_of_color, 0, 0))
    # 3. 初始化画布,并添加结点和Child边
    nt = Network('1000px', '1000px', directed=True)
    nt.add_nodes(range(node_nums),
                 title=[str(x) for x in range(node_nums)],
                 label=[one_data['ContextGraph']['NodeLabels'][str(x)] for x in range(node_nums)],
                 color=color
                 )
    nt.add_edges(child_edges)
    # 4. 依次添加其他需要额外显示的边
    for idx, (label) in enumerate(labels):
        try:
            to_show_edges = one_data['ContextGraph']['Edges'][label]
            for edge in to_show_edges:
                nt.add_edge(edge[0], edge[1],
                            value=4,
                            label=label)
        except:
            print('no ' + label + ' edges')
            return
    # 5. 显示图
    nt.show(path.split('/')[-1].split('.')[0] + str(index) + '.html')

Be First to Comment

发表回复

您的电子邮箱地址不会被公开。