机器学习——AdaBoosting
Posted 颜妮儿
tags:
篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了机器学习——AdaBoosting相关的知识,希望对你有一定的参考价值。
AdaBoosting算法是一种集成算法。
集成算法是通过构建并结合多个学习器来完成学习任务,就是“三个臭皮匠赛过诸葛亮”的思想。
例如,我们现有错误率为0.45的弱分类器,假设各个分类器相互独立,则集成后的错误率随个体分类器的数目关系为:
不难发现,随着个体分类器数目增大,集成后的错误率随之下降。
结合Hoeffding不等式的关系表达式为:
P
(
F
(
x
)
≠
f
(
x
)
)
=
∑
k
=
0
⌊
T
/
2
⌋
(
T
k
)
(
1
−
ϵ
)
k
ϵ
T
−
k
≤
e
−
1
2
T
(
1
−
2
ϵ
)
2
P(F(\\boldsymbolx)\\ne f(\\boldsymbolx))=\\sum\\limits_k=0^\\lfloor T/2\\rfloor \\beginpmatrix T\\\\ k \\endpmatrix (1-\\epsilon)^k\\epsilon^T-k\\le e^-\\frac12T(1-2\\epsilon)^2
P(F(x)=f(x))=k=0∑⌊T/2⌋(Tk)(1−ϵ)kϵT−k≤e−21T(1−2ϵ)2
其中
T
T
T表示集成中个体分类器的个数;
ϵ
\\epsilon
ϵ表示个体分类器的错误率;
F
(
x
)
和
f
(
x
)
F(\\boldsymbol x)和f(\\boldsymbolx)
F(x)和f(x)分别表示预测标签和真实标签。
不难从式子中发现:集成后的错误率随个体分类器的个数增大而指数级下降;当个体分类器的错误率大于等于0.5时,集成没有意义。
关于个体分类器的权值设置和分类错误的样本的权重设置的更新表达式:
这两个式子是由最小化指数损失函数推导而来的,具体的推导过程可见西瓜书集成学习这一节内容。
当前个体分类器的权重:
α
=
1
2
ln
(
1
−
ϵ
ϵ
)
\\alpha=\\frac12\\ln\\left(\\frac1-\\epsilon\\epsilon\\right)
α=21ln(ϵ1−ϵ),由于当错误率
ϵ
\\epsilon
ϵ大于等于0.5时没有意义,所以
α
>
0
\\alpha\\gt 0
α>0;
更新错误样本的权重:
D
t
+
1
(
x
)
=
D
t
(
x
)
Z
t
×
e
−
α
f
(
x
)
h
(
x
)
D_t+1(\\boldsymbolx)=\\fracD_t(\\boldsymbolx)Z_t\\times e^-\\alpha f(\\boldsymbol x)h(\\boldsymbol x)
Dt+1(x)=ZtDt(x)×e−αf(x)h(x)(当
f
(
x
)
=
h
(
x
)
f(\\boldsymbol x)=h(\\boldsymbol x)
f(x)=h(x)时,
f
(
x
)
h
(
x
)
=
1
f(\\boldsymbol x)h(\\boldsymbol x)=1
f(x)h(x)=1,否则为-1。当预测值和真实标签不同时,该样本权重为增大,否则会减少),
Z
t
Z_t
Zt是规范化因子,将权重缩放到0~1。
关于权重的理解:
为什么要增大预测错误的样本权重?结合计算平均错误率公式: E ( f ; D ) = 1 m ∑ i = 1 m I ( f ( x i ) ≠ y i ) E(f;D)=\\frac1m\\sum\\limits_i=1^m\\mathbbI(f(\\boldsymbol x_i)\\ne y_i) E(f;D)=m1i=1∑mI(f(xi)=yi),如果样本不是均匀分布,例如样本属性值为 x k x_k xk的样本出现次数为 k k k,则该样本的出现概率表示为: p ( x k ) = k m p(\\boldsymbol x_k)=\\frackm p(xk)=mk,则平均错误率表示为: E ( f ; D ) = ∫ x ∼ D I ( f ( x i ) ≠ y i ) p ( x ) d x E(f;D)=\\int_\\boldsymbol x\\sim D \\mathbbI(f(\\boldsymbol x_i)\\ne y_i)p(\\boldsymbol x)dx E(f;D)=∫x∼DI(f(xi)=yi)p(x)dx,增大其样本权重相当于增加了该样本出现的频率,频率越大,则对该样本的分类效果越好。
为什么要给个体分类器设置权重?在集合后的分类根据每个个体分类器的预测结果投票获得最终结果,根据个体分类器的权重计算表达式, 1 − ϵ ϵ \\frac1-\\epsilon\\epsilon ϵ1−ϵ表示分类器分类正确的几率,值越大则表示该分类器的效果越好,它在投票时的话语权也应该越高。
代码:
(1)
package adaboosting;
import java.io.FileReader;
import java.util.Arrays;
import weka.core.Instances;
public class WeightedInstances extends Instances
/**
* Just the require of some classes, any number is OK.
*/
private static final long serialVersionUID = 11087456L;
/**
* Weights.
*/
private double[] weights;
/**
*********************
* The first constructor.
*
* @param paraFileReader The given reader to read data from file.
*********************
*/
public WeightedInstances(FileReader paraFileReader) throws Exception
super(paraFileReader);
setClassIndex(numAttributes() - 1);
// Initialize weights
weights = new double[numInstances()];
double tempAverage = 1.0 / numInstances();
for (int i = 0; i < weights.length; i++)
weights[i] = tempAverage;
// Of for i
System.out.println("Instances weights are: " + Arrays.toString(weights));
// Of the first constructor.
/**
*********************
* The second constructor.
*
* @param paraInstances The given instance.
*********************
*/
public WeightedInstances(Instances paraInstances)
super(paraInstances);
setClassIndex(numAttributes() - 1);
// Initialize weights.
weights = new double[numInstances()];
double tempAverage = 1.0 / numInstances();
for (int i = 0; i < weights.length; i++)
weights[i] = tempAverage;
// Of for i
System.out.println("Instances weights are: " + Arrays.toString(weights));
// Of the second constructor
/**
********************
* Getter.
*
* @param paraIndex The given index.
* @return The weight of the given index.
*********************
*/
public double getWeight(int paraIndex)
return weights[paraIndex];
// Of getWeight
/**
********************
* Adjust the weights.
*
* @param paraCorrectArray Indicate which instances have been correctly
* classified.
* @param paraAlpha The weight of the last classifier.
*********************
*/
public void adjustWeights(boolean[] paraCorrectArray, double paraAlpha)
// Step 1. Calculate alpha.
double tempIncrease = Math.exp(paraAlpha);
// Step 2. Adjust
double tempWeightsSum = 0;// For normalization
for (int i = 0; i < weights.length; i++)
if (paraCorrectArray[i])
weights[i] /= tempIncrease;
else
weights[i] *= tempIncrease;
// Of if
tempWeightsSum += weights[i];
// Of for i
// Step 3. Normalize.
for (int i = 0; i < weights.length; i++)
weights[i] /= tempWeightsSum;
System.out.println("After adjusting, instances weights are: " + Arrays.toString(weights));
// Of adjustWeights
/**
********************
* Test the method.
*********************
*/
public void adjustWeightsTest()
boolean[] tempCorrectArray = new boolean[numInstances()];
for (int i = 0; i < tempCorrectArray.length; i++)
tempCorrectArray[i] = true;
// Of for i
double tempWeightedError = 0.3;
adjustWeights(tempCorrectArray, tempWeightedError);
System.out.println("After adjusting");
System.out.println(toString());
// Of adjustWeightsTest
/**
*********************
* For display.
*********************
*/
public String toString()
String resultString = "I am a weighted Instances object.\\r\\n" + "I have " + numInstances() + " instances and "
+ (numAttributes() - 1) + " conditional attributes.\\r\\n" + "My weights are: " + Arrays.toString(weights)
+ "\\r\\n" + "My data are: " + super.toString();
return resultString;
// Of toString
/**
********************
* For unit test.
*
* @param args Not provided.
*********************
*/以上是关于机器学习——AdaBoosting的主要内容,如果未能解决你的问题,请参考以下文章