com.hankcs.hanlp.model.perceptron.model

Class LinearModel

java.lang.Object

com.hankcs.hanlp.model.perceptron.model.LinearModel

All Implemented Interfaces:

ICacheAble

Direct Known Subclasses:

AveragedPerceptron, LogLinearModel, StructuredPerceptron

public class LinearModel
extends Object
implements ICacheAble

在线学习标注模型

Author:

hankcs

Field Summary

Fields

Modifier and Type	Field and Description
FeatureMap	featureMap 特征函数
float[]	parameter 特征权重

Constructor Summary

Constructors

Constructor and Description
LinearModel(FeatureMap featureMap)
LinearModel(FeatureMap featureMap, float[] parameter)
LinearModel(String modelFile)

Method Summary

Modifier and Type	Method and Description
LinearModel	compress(double ratio) 模型压缩
LinearModel	compress(double ratio, double threshold)
int	decode(Collection<Integer> x) 分离超平面解码
boolean	load(ByteArray byteArray) 载入
void	load(String modelFile) 加载模型
void	save(DataOutputStream out) 写入
void	save(String modelFile) 保存到路径
void	save(String modelFile, double ratio) 压缩并保存
void	save(String modelFile, Set<Map.Entry<String,Integer>> featureIdSet, double ratio)
void	save(String modelFile, Set<Map.Entry<String,Integer>> featureIdSet, double ratio, boolean text) 保存
double	score(int[] featureVector, int currentTag) 通过命中的特征函数计算得分
TagSet	tagSet()
TaskType	taskType()
void	update(Collection<Integer> x, int y) 参数更新
double	viterbiDecode(Instance instance) 维特比解码
double	viterbiDecode(Instance instance, int[] guessLabel) 维特比解码

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

featureMap

public FeatureMap featureMap

特征函数

parameter

public float[] parameter

特征权重

Constructor Detail

LinearModel

public LinearModel(FeatureMap featureMap,
                   float[] parameter)

LinearModel

public LinearModel(FeatureMap featureMap)

LinearModel

public LinearModel(String modelFile)
            throws IOException

Throws:

IOException

Method Detail

compress

public LinearModel compress(double ratio)

模型压缩

Parameters:

ratio - 压缩比c（压缩掉的体积，压缩后体积变为1-c）

Returns:

compress

public LinearModel compress(double ratio,
                            double threshold)

Parameters:

ratio - 压缩比c（压缩掉的体积，压缩后体积变为1-c）

threshold - 特征权重绝对值之和最低阈值

Returns:

save

public void save(String modelFile)
          throws IOException

保存到路径

Parameters:

modelFile -

Throws:

IOException

save

public void save(String modelFile,
                 double ratio)
          throws IOException

压缩并保存

Parameters:

modelFile - 路径

ratio - 压缩比c（压缩掉的体积，压缩后体积变为1-c）

Throws:

IOException

save

public void save(String modelFile,
                 Set<Map.Entry<String,Integer>> featureIdSet,
                 double ratio)
          throws IOException

Throws:

IOException

save

public void save(String modelFile,
                 Set<Map.Entry<String,Integer>> featureIdSet,
                 double ratio,
                 boolean text)
          throws IOException

保存

Parameters:

modelFile - 路径

featureIdSet - 特征集（有些数据结构不支持遍历，可以提供构造时用到的特征集来规避这个缺陷）

ratio - 压缩比

text - 是否输出文本以供调试

Throws:

IOException

update

public void update(Collection<Integer> x,
                   int y)

参数更新

Parameters:

x - 特征向量

y - 正确答案

decode

public int decode(Collection<Integer> x)

分离超平面解码

Parameters:

x - 特征向量

Returns:

sign(wx)

viterbiDecode

public double viterbiDecode(Instance instance)

维特比解码

Parameters:

instance - 实例

Returns:

viterbiDecode

public double viterbiDecode(Instance instance,
                            int[] guessLabel)

维特比解码

Parameters:

instance - 实例

guessLabel - 输出标签

Returns:

score

public double score(int[] featureVector,
                    int currentTag)

通过命中的特征函数计算得分

Parameters:

featureVector - 压缩形式的特征id构成的特征向量

Returns:

load

public void load(String modelFile)
          throws IOException

加载模型

Parameters:

modelFile -

Throws:

IOException

tagSet

public TagSet tagSet()

save

public void save(DataOutputStream out)
          throws IOException

Description copied from interface: ICacheAble

写入

Specified by:

save in interface ICacheAble

Throws:

IOException

load

public boolean load(ByteArray byteArray)

Description copied from interface: ICacheAble

载入

Specified by:

load in interface ICacheAble

Returns:

taskType

public TaskType taskType()