GNN2.1 DeepWalk (Code-Implementation)

GNN学习笔记

GNN从入门到精通课程笔记

2.1 DeepWalk (Code-Implementation)

DeepWalk: Online Learning of Social Representations (KDD’ 14)

import networkx as nx
import pandas as pd
import numpy as np
import random
from tqdm import tqdm
from gensim.models import Word2Vec

df = pd.read_csv("dataset/seealsology-data.tsv", sep = "\t")
print(df.head())

# Create a graph
G = nx.from_pandas_edgelist(df, source = "source", target = "target", edge_attr = True, create_using = nx.Graph())
print("Number of nodes:", G.number_of_nodes())
print("Number of edges:", G.number_of_edges())

def random_walk(G, node, walk_length):
    """
    Simulate a random walk starting from start node.
    G: NetworkX graph
    node: Start node for the random walk
    walk_length: Length of the random walk
    """
    
    walk = [node]
    for i in range(walk_length - 1):
        neighbors = list(G.neighbors(node))
        if len(neighbors) == 0:
            break
        node = random.choice(neighbors)
        walk.append(node)
    return walk

# Generate walks
all_nodes = list(G.nodes())
random_walks = []

for n in tqdm(all_nodes):
    for i in range(5):
        random_walks.append(random_walk(G, n, 10))
print("Number of random walks:", len(random_walks))

# Train Word2Vec model
model = Word2Vec(window = 4, sg = 1, hs = 0,
                 negative=10, # for negative sampling
                 alpha=0.03, min_alpha=0.0007,
                 seed=14)

model.build_vocab(random_walks, progress_per=2)
model.train(random_walks, total_examples = model.corpus_count, epochs=20, report_delay=1)
print(model)

运行结果

           source                         target  depth
0  space research     advances in space research      1
1  space research  benefits of space exploration      1
2  space research    committee on space research      1
3  space research         deep space exploration      1
4  space research        lists of space programs      1
Number of nodes: 510
Number of edges: 680
100%|████████████████████████████████████████| 510/510 [00:00<00:00, 15529.95it/s]
Number of random walks: 2550
Word2Vec<vocab=510, vector_size=100, alpha=0.03>