向量最近邻训练 (VectorNearestNeighborTrainBatchOp)

Java 类名：com.alibaba.alink.operator.batch.similarity.VectorNearestNeighborTrainBatchOp

Python 类名：VectorNearestNeighborTrainBatchOp

功能介绍

该组件为向量最近邻的训练过程，在计算时与 VectorNearestNeighborPredictBatchOp 配合使用。

支持的距离计算方式包含EUCLIDEAN，COSINE，INNERPRODUCT（内积），CITYBLOCK（曼哈顿距离），JACCARD，PEARSON

默认距离EUCLIDEAN

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
idCol	id列名	id列名	String	✓
selectedCol	选中的列名	计算列对应的列名	String	✓	所选列类型为 [DENSE_VECTOR, SPARSE_VECTOR, STRING, VECTOR]
metric	距离度量方式	聚类使用的距离类型	String		“EUCLIDEAN”, “COSINE”, “INNERPRODUCT”, “CITYBLOCK”, “JACCARD”, “PEARSON”	“EUCLIDEAN”

代码示例

Python 代码

from pyalink.alink import *

import pandas as pd

useLocalEnv(1)

df = pd.DataFrame([
    [0, "0 0 0"],
    [1, "1 1 1"],
    [2, "2 2 2"]
])

inOp = BatchOperator.fromDataframe(df, schemaStr='id int, vec string')
train = VectorNearestNeighborTrainBatchOp().setIdCol("id").setSelectedCol("vec").linkFrom(inOp)
predict = VectorNearestNeighborPredictBatchOp().setSelectedCol("vec").setTopN(3).linkFrom(train, inOp)
predict.print()

Java 代码

import org.apache.flink.types.Row;

import com.alibaba.alink.operator.batch.BatchOperator;
import com.alibaba.alink.operator.batch.similarity.VectorNearestNeighborPredictBatchOp;
import com.alibaba.alink.operator.batch.similarity.VectorNearestNeighborTrainBatchOp;
import com.alibaba.alink.operator.batch.source.MemSourceBatchOp;
import org.junit.Test;

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

public class VectorNearestNeighborTrainBatchOpTest {
	@Test
	public void testVectorNearestNeighborTrainBatchOp() throws Exception {
		List <Row> df = Arrays.asList(
			Row.of(0, "0 0 0"),
			Row.of(1, "1 1 1"),
			Row.of(2, "2 2 2")
		);
		BatchOperator <?> inOp = new MemSourceBatchOp(df, "id int, vec string");
		BatchOperator <?> train =
			new VectorNearestNeighborTrainBatchOp().setIdCol("id").setSelectedCol("vec").linkFrom(
			inOp);
		BatchOperator <?> predict =
			new VectorNearestNeighborPredictBatchOp().setSelectedCol("vec").setTopN(3).linkFrom(
			train, inOp);
		predict.print();
	}
}

运行结果

id	vec
0	{“ID”:“[0,1,2]”,“METRIC”:“[0.0,1.7320508075688772,3.4641016151377544]”}
1	{“ID”:“[1,2,0]”,“METRIC”:“[0.0,1.7320508075688772,1.7320508075688772]”}
2	{“ID”:“[2,1,0]”,“METRIC”:“[0.0,1.7320508075688772,3.4641016151377544]”}