大规模DeepWalk (HugeDeepWalkTrainBatchOp)

Java 类名：com.alibaba.alink.operator.batch.huge.HugeDeepWalkTrainBatchOp

Python 类名：HugeDeepWalkTrainBatchOp

功能介绍

DeepWalk是2014年提出的一个新的方法，用来为网络中的结点学习隐式特征表达，即将网络中的每一个点表示成连续特征空间中的一个点向量。DeepWalk是无监督特征学习方法，利用随机游走(Random Walk)及语言模型(Language modeling)，学习出的隐式特征能够捕捉到网络的结构信息。后续论文也提出了一些扩展，如结合损失函数进行有监督学习、结合文本信息等等

DeepWalk: Online Learning of Social Representations

参数说明

名称	中文名称	描述	类型	是否必须？	取值范围	默认值
sourceCol	起始点列名	用来指定起始点列	String	✓
targetCol	中止点点列名	用来指定中止点列	String	✓
walkLength	游走的长度	随机游走完向量的长度	Integer	✓
walkNum	路径数目	每一个起始点游走出多少条路径	Integer	✓
alpha	学习率	学习率	Double			0.025
batchSize	batch大小	batch大小, 按行计算	Integer		x >= 1
isToUndigraph	是否转无向图	选为true时，会将当前图转成无向图，然后再游走	Boolean			false
minCount	最小词频	最小词频	Integer			5
negative	负采样大小	负采样大小	Integer			5
numCheckpoint	checkPoint 数目	checkPoint 数目	Integer			1
numIter	迭代次数	迭代次数，默认为1。	Integer			1
randomWindow	是否使用随机窗口	是否使用随机窗口，默认使用	String			“true”
vectorSize	embedding的向量长度	embedding的向量长度	Integer		x >= 1	100
weightCol	权重列名	权重列对应的列名	String		所选列类型为 [BIGDECIMAL, BIGINTEGER, BYTE, DOUBLE, FLOAT, INTEGER, LONG, SHORT]	null
window	窗口大小	窗口大小	Integer			5

代码示例

Python 代码

from pyalink.alink import *

import pandas as pd

useLocalEnv(1)

df_data = pd.DataFrame([
    ["Bob", "Lucy", 1.],
    ["Lucy", "Bob", 1.],
    ["Lucy", "Bella", 1.],
    ["Bella", "Lucy", 1.],
    ["Alice", "Lisa", 1.],
    ["Lisa", "Alice", 1.],
    ["Lisa", "Karry", 1.],
    ["Karry", "Lisa", 1.],
    ["Karry", "Bella", 1.],
    ["Bella", "Karry", 1.]
])
source =  BatchOperator.fromDataframe(df_data, schemaStr='start string, end string, value double')

deepWalkBatchOp = HugeDeepWalkTrainBatchOp() \
  .setSourceCol("start")            \
  .setTargetCol("end")              \
  .setWeightCol("value")            \
  .setWalkNum(2)                    \
  .setWalkLength(2)                 \
  .setMinCount(1)                   \
  .setVectorSize(4)
deepWalkBatchOp.linkFrom(source).print()

Java 代码

import org.apache.flink.types.Row;

import com.alibaba.alink.operator.batch.BatchOperator;
import com.alibaba.alink.operator.batch.huge.HugeDeepWalkTrainBatchOp;
import com.alibaba.alink.operator.batch.source.MemSourceBatchOp;
import org.junit.Test;

import java.util.Arrays;
import java.util.List;

public class HugeDeepWalkTrainBatchOpTest {
	@Test
	public void testHugeDeepWalkTrainBatchOp() throws Exception {
		List <Row> df_data = Arrays.asList(
			Row.of("Bob", "Lucy", 1.),
			Row.of("Lucy", "Bob", 1.),
			Row.of("Lucy", "Bella", 1.),
			Row.of("Bella", "Lucy", 1.),
			Row.of("Alice", "Lisa", 1.),
			Row.of("Lisa", "Alice", 1.),
			Row.of("Lisa", "Karry", 1.),
			Row.of("Karry", "Lisa", 1.),
			Row.of("Karry", "Bella", 1.),
			Row.of("Bella", "Karry", 1.)
		);
		BatchOperator <?> source = new MemSourceBatchOp(df_data, "start string, end string, value double");
		BatchOperator <?> deepWalkBatchOp = new HugeDeepWalkTrainBatchOp()
			.setSourceCol("start")
			.setTargetCol("end")
			.setWeightCol("value")
			.setWalkNum(2)
			.setWalkLength(2)
			.setMinCount(1)
			.setVectorSize(4);
		deepWalkBatchOp.linkFrom(source).print();
	}
}

运行结果

node	vec
Karry	0.03438692167401314,-0.04779096320271492,0.012648836709558964,-0.09576538950204849
Lisa	0.11595723778009415,-0.08507091552019119,0.1099027618765831,0.013517010025680065
Bella	0.05783883109688759,0.08286115527153015,-0.06497485190629959,0.026532595977187157
Alice	0.05775630846619606,-0.099935382604599,-0.022451162338256836,-0.023144230246543884
Lucy	0.11699658632278442,0.05271214246749878,-0.12347490340471268,-0.08684996515512466
Bob	-0.07306862622499466,-0.11596906185150146,-0.04183155298233032,0.03973118215799332

ALinkLab