图神经网络-图采样Graphsage代码实现

一：为什么要图采样？

二 Graphsage 采样代码实践

GraphSage的PGL完整代码实现位于https://github.com/PaddlePaddle/PGL/tree/main/examples/graphsage，本文实现一个简单的graphsage 采样代码 。

安装依赖

# !pip install paddlepaddle==1.8.4
!pip install pgl -q

1. 构建graph

图网络的构建使用Graph类，Graph类的具体实现可以参考https://github.com/PaddlePaddle/PGL/blob/main/pgl/graph.py

import random
import numpy as np
import pgl
import display


def build_graph():
    # 定义节点的个数；每个节点用一个数字表示，即从0~9
    num_node = 16
    # 添加节点之间的边，每条边用一个tuple表示为: (src, dst)
    edge_list = [(2, 0), (1, 0), (3, 0),(4, 0), (5, 0), 
             (6, 1), (7, 1), (8, 2), (9, 2), (8, 7),
             (10, 3), (4, 3), (11, 10), (11, 4), (12, 4),
             (13, 5), (14, 5), (15, 5)]


    g = pgl.graph.Graph(num_nodes = num_node, edges = edge_list)


    return g


# 创建一个图对象，用于保存图网络的各种数据。
g = build_graph()
display.display_graph(g)

运行结果：

2. GraphSage采样函数实现

GraphSage的作者提出采样算法来使得模型能够以Mini-batch的方式进行训练，算法代码见论文附录A https://cs.stanford.edu/people/jure/pubs/graphsage-nips17.pdf。

• 假设要利用中心节点的k阶邻居信息，则在聚合的时候，需要从第k阶邻居传递信息到k-1阶邻居，并依次传递到中心节点。

• 采样的过程与此相反，在构造第t轮训练的Mini-batch时，从中心节点出发，在前序节点集合中采样Nt个邻居节点加入采样集合。

• 将邻居节点作为新的中心节点继续进行第t-1轮训练的节点采样，以此类推。

• 将采样到的节点和边一起构造得到子图。

def traverse(item):
    """traverse
    """
    if isinstance(item, list) or isinstance(item, np.ndarray):
        for i in iter(item):
            for j in traverse(i):
                yield j
    else:
        yield item


def flat_node_and_edge(nodes):
    """这个函数的目的是为了将 list of numpy array 扁平化成一个list
    例如：[array([7, 8, 9]), array([11, 12]), array([13, 15])] --> [7, 8, 9, 11, 12, 13, 15]
    """
    nodes = list(set(traverse(nodes)))
    return nodes


def graphsage_sample(graph, start_nodes, sample_num):
    subgraph_edges = []
    # pre_nodes: a list of numpy array, 
    pre_nodes = graph.sample_predecessor(start_nodes, sample_num)


    # 根据采样的子节点， 恢复边
    for dst_node, src_nodes in zip(start_nodes, pre_nodes):
        for node in src_nodes:
            subgraph_edges.append((node, dst_node))




    subgraph_nodes = flat_node_and_edge(pre_nodes)


    return subgraph_nodes, subgraph_edges

随机获取一阶邻居信息

seed = 458
np.random.seed(seed)
random.seed(seed)


start_nodes = [0]


layer1_nodes, layer1_edges = graphsage_sample(g, start_nodes, sample_num=3)
print('layer1_nodes: ', layer1_nodes)
print('layer1_edges: ', layer1_edges)
display.display_subgraph(g, {'orange': layer1_nodes}, {'orange': layer1_edges})

运行结果

layer1_nodes:  [2, 4, 5]
layer1_edges:  [(4, 0), (2, 0), (5, 0)]

继续获取二阶邻居节点信息

layer2_nodes, layer2_edges = graphsage_sample(g, layer1_nodes, sample_num=2)
print('layer2_nodes: ', layer2_nodes)
print('layer2_edges: ', layer2_edges)
display.display_subgraph(g, {'orange': layer1_nodes, 'Thistle': layer2_nodes}, {'orange': layer1_edges, 'Thistle': layer2_edges})

运行结果

layer2_nodes:  [8, 9, 11, 12, 14, 15]
layer2_edges:  [(8, 2), (9, 2), (11, 4), (12, 4), (14, 5), (15, 5)]

图节点可视化代码

#%matplotlib inline
import matplotlib.pyplot as plt
import numpy as np
import networkx as nx # networkx是一个常用的绘制复杂图形的Python包。


def display_graph(g):
    nx_G = nx.Graph()
    nx_G.add_nodes_from(range(g.num_nodes))
    nx_G.add_edges_from(g.edges)


    pos = {0: [0.5, 0.5], 1:[0.6, 0.4], 2:[0.47, 0.67], 3: [0.35, 0.55], 4:[0.4, 0.4], 5:[0.5, 0.3],
           6: [0.8, 0.4], 7:[0.65, 0.65], 8:[0.6, 0.8], 9:[0.45, 0.85], 10:[0.15, 0.7], 11: [0.1, 0.4],
           12:[0.2, 0.2], 13:[0.3, 0.1], 14:[0.55, 0.15], 15:[0.7, 0.22]}
    nx.draw(nx_G, 
            pos,
            with_labels=True,
            node_color='green', 
            edge_color='green',
            node_size=1000)


    plt.show()


#display_graph(g)# 创建一个GraphWrapper作为图数据的容器，用于构建图神经网络。


def display_subgraph(g, sub_nodes, sub_edges):
    nx_G = nx.Graph()
    nx_G.add_nodes_from(range(g.num_nodes))
    nx_G.add_edges_from(g.edges)


    pos = {0: [0.5, 0.5], 1:[0.6, 0.4], 2:[0.47, 0.67], 3: [0.35, 0.55], 4:[0.4, 0.4], 5:[0.5, 0.3],
           6: [0.8, 0.4], 7:[0.65, 0.65], 8:[0.6, 0.8], 9:[0.45, 0.85], 10:[0.15, 0.7], 11: [0.1, 0.4],
           12:[0.2, 0.2], 13:[0.3, 0.1], 14:[0.55, 0.15], 15:[0.7, 0.22]}
    nx.draw(nx_G, 
            pos,
            with_labels=True,
            node_color='green',
            edge_color='green',
            node_size=1000,
            width=1)


    nx.draw_networkx_nodes(nx_G, pos, nodelist=[0], node_color='red', node_size=1000)


    for color, nodes in sub_nodes.items():
        nx.draw_networkx_nodes(nx_G, pos, nodelist=nodes, node_color=color, node_size=1000)


    for color, edges in sub_edges.items():
        nx.draw_networkx_edges(nx_G, pos, edgelist=edges, edge_color=color, width=5)


    plt.show()

注：本文图文资料来源于 AIStudio-人工智能学习与实训社区