Chapter 4 Introduction of Boosting

分类是机器学习的基本问题，提高分类器的泛化性能是分类器研究的一个主要目标
Classification is the basic problem of machine learning to improve the generalization performance of classifier is classifier a major objective.

分类方法有很多，如支持向量机、决策树、贝叶斯方法和神经网络等，但是目前主流的三种分类方法包括SVM(支持向量机)，Boosting(主要是AdaBoost)，和Logistic Regression。

There are many classification methods, such as support vector machines, decision trees, Bayesian methods and neural networks, but current classification methods include the three main SVM (support vector machine), Boosting (mainly AdaBoost), and Logistic Regression.

其中Boosting作为一种新的集成机器学习方法，以学习理论为依据，在很多应用领域都表现出了优良特性。

Boosting them as a new integrated machine learning approach to learning theory as the basis, in many applications have shown excellent characteristics.

4.1 Boosting Algorithm background （4.1 Boosting算法的背景）

Boosting是一种试图提升任意给定学习算法精度的普遍方法。

Boosting is an attempt to enhance the accuracy of any given learning algorithm for the general approach.

它的思想起源于Valiant提出的计算学习理论--PAC(Probably Approximately Correct)学习模型。

It put forward the idea originated in the calculation of Valiant learning theory - PAC (Probably Approximately Correct) learning model.

PAC(概率近似正确)是统计机器学习、集成机器学习等方法的理论基础。

PAC (probability near the right) is a statistical machine learning, machine learning and other methods of integration theory.

Keams和Valiant首先提出下面问题：在Valiant的PAC模型中，一个性能仅比随机猜测稍好的“弱”学习算法是否能被“提升”为一个具有任意精度的“强”学习算法?

Keams and Valiant first put forward the following question: In Valiant's PAC model, a performance only slightly better than random guessing "weak" learning algorithm it can be "upgrading" to an arbitrary precision of "strong" learning algorithm?

1990年Schapire提出了第一个可证明的多项式时间Boosting算法，对这个闯题做出了肯定的回答。

1990, Schapire proposed the first provable polynomial time Boosting algorithms, Chuang title of this made an affirmative answer.

Schapire证明，如果将多个PAC分类器集成在一起，它将具有PAC强分类器的泛化能力。

Schapire proved that, if the number of integrated PAC classifier, it will have PAC strong classifier generalization.

进而又说明，这类集成后的强分类器具有统计学习理论的基础。之后，Freund设计了一个更加高效的通过重取样或过滤运作的Boost-by-majority算法。

In turn, shows that after such integration has strong classifier based on statistical learning theory.

之后，Freund设计了一个更加高效的通过重取样或过滤运作的Boost-by-majority算法。

Later, Freund designed a more efficient re-sampling or filtering through the operation of the Boost-by-majority algorithm.

这个算法尽管在某种意义上是优化的，但却有一些实践上的缺陷。1995年Freund与Schapire提出AdaBoost(Adaptive Boosting)算法。

Although this algorithm is optimal in some sense, but there are some practical deficiencies. 1995, Freund and Schapire proposed AdaBoost (Adaptive Boosting) algorithm.

这个算法和“Boost-by-majority”算法的效率几乎一样，却可以非常容易的应用到实际问题中去。
The algorithm and "Boost-by-majority" algorithm is nearly the same efficiency, but can be easily applied to practical problems.

然后，Boosting算法经过进一步改进又有了很大的发展，如通过调整权重而运作的AdaBoost．M1、AdaBoost．M2、AdaBoost．R算法等，解决了早期的Boosting算法很多实践上的问题。

Then, Boosting algorithm is further improved, there has been a significant development, such as by adjusting the weights and the operation of AdaBoost. M1, AdaBoost. M2, AdaBoost. R algorithm to solve many of the early Boosting algorithm practical problems.

Chapter 4 Introduction of Boosting
Classification is the basic problem of machine learning to improve the generalization performance of classifier is classifier a major objective. There are many classification methods, such as support vector machines, decision trees, Bayesian methods and neural networks, but current classification methods include the three main SVM (support vector machine), Boosting (mainly AdaBoost), and Logistic Regression. Boosting them as a new integrated machine learning approach to learning theory as the basis, in many applications have demonstrated the superior characteristics.
4.1 Boosting Algorithm background
Boosting is an attempt to enhance the accuracy of any given learning algorithm for the general approach. It put forward the idea originated in the calculation of Valiant learning theory - PAC (Probably Approximately Correct) learning model. PAC (probability approximately correct) is the statistical machine learning, machine learning and other methods of integration theory.
Keams and Valiant first put forward the following question: In Valiant's PAC model, a performance only slightly better than random guessing "weak" learning algorithm it can be "upgrading" to an arbitrary precision of "strong" learning algorithm?
1990, Schapire proposed first provable polynomial time Boosting algorithm, the Chuang made a positive answer questions. Schapire proved that, if the number of integrated PAC classifier, it will have strong classifier PAC generalization ability. In turn, shows that after such integration has strong classifier based on statistical learning theory. Later, Freund designed a more efficient re-sampling or filtering through the operation of the Boost-by-majority algorithm. Although this algorithm is optimal in some sense, but there are some practical deficiencies.
Schapire 1995, Freund made with the AdaBoost (Adaptive Boosting) algorithm.
The algorithm and "Boost-by-majority" algorithm is nearly the same efficiency, but can be easily applied to practical problems. Then, Boosting algorithm is further improved, there has been a significant development, such as by adjusting the weights and the operation of AdaBoost. M1, AdaBoost. M2, AdaBoost. R algorithm to solve many of the early Boosting algorithm practical problems.

Chapter 4 Introduction of Boosting

Classification is the basic problem of machine learning, improve classifier generalization capability is one of the main classifier research goal. Many methods, such as classification of support vector machines, the decision tree, the bayesian method and neural networks, but the mainstream of the three types of classification methods including SVM (support vector machine), Boosting (mainly Logistic Regression, and AdaBoost). As a new kind of Boosting of machine learning method, integrated learning theory basis, in many applications show the excellent characteristics.

Boosting algorithm background. 4.1

Boosting is an attempt to ascend any given learning algorithm of common methods. Precision, It originated from the thought of learning theory Valiant calculation of temperature (Approximately - PAC) learning model. PAC (right) are approximate probability statistics, machine learning, integration of machine learning method of theoretical basis.
Keams Valiant and put forward the following questions: first, the PAC Valiant model in a performance than random speculation only slightly better "weak" learning algorithm can be "ascension" for an arbitrary precision of the "strong" learning algorithms?
1990 first Schapire proposed to prove that the algorithm of Boosting polynomial time rushing made sure the problem of answer. Schapire proof, if the PAC classifier integrated together, it will have a PAC strong classifier generalization ability. Then again, this kind of strong after integration has a statistical learning theory basis. After a more efficient Freund design through heavy or filtering operation to hee majority algorithm. Despite this algorithm, in a sense, is optimized, but there are some defects in practice. In 1995 Schapire with AdaBoost proposed Freund Adaptive Boosting) algorithm (.
This algorithm and the "to" hee majority efficiency of the algorithm is almost the same, but can be easily applied to practical problems. Then, after further improvement and Boosting algorithm is developed, such as through adjusting weight and operation of M1 AdaBoost. J, AdaBoost algorithm AdaBoost. R M2, etc, to solve the early Boosting algorithm on the practice of many problems.

(乌哈习14729098500)高分悬赏,英语高手请进!翻译论文摘要,拒绝翻译软件翻译,主要是翻译软件直译错误多!十分的谢谢!... ______ 希望对你有帮助! 翻译答案是: Wound cascade induction motor speed control system design Abstract Speed is the cascade-type motor rotor winding loop of string attached to the adjustable potential to change the motor to the poor, to achieve ...

(乌哈习14729098500)帮忙翻译一下论文摘要,悬赏5分以作答谢! - ______ Abstract: VPN (Virtual Private Network) that the virtual private network, VPN is the expansion of enterprise network, it can help by remote users, the company branches, commercial partners and suppliers with the company's internal network linking ...

(乌哈习14729098500)请帮我翻译一下论文摘要,翻译的好,翻译的专业有悬赏分. - ______ This article analyses the speed, efficiency, advantages and disadvantages by using different drawing draft method.A centimeter-level 3D positioning achievement provided by a factor based on the RTK real time dynamic GPS measuring technology...

(乌哈习14729098500)求论文翻译(题目,摘要,关键词)悬赏100分 - ______ Title:Wanxi internship Educational Institute of Problems and Solutions Abstract: Education teacher training institutions are training students to teachers the importance of the process of transformation.Wanxi University as a result of teacher training ...

(乌哈习14729098500)翻译一篇完形填空``悬赏10分`` ______ 一天,爸爸和我一起去马戏团看表演.在我们前边有一个家庭也在排队买票. 这个家庭共有8个孩子.他们穿的衣服并不是很昂贵但是却十分很干净.他们正在激动的谈论着小丑和大象. “我买10张票,8个孩子的、2个大人的.”他们的爸爸...

(乌哈习14729098500)谁能帮我把这篇小论文翻译成英文!!!!有悬赏分啊啊啊~~急.谢谢!!!英文英文英文翻译! - ______ Chinese eco-agriculture theory and industrializationtheEcological agriculture industrialization is a new period of agricultural development direction, agriculture, farmers' income is the effective way. China should establish the conesponding ...

(乌哈习14729098500)翻译一篇完形填空``悬赏10分``Oneday,aboyhadafightwithoneofhisclassmates.Thenhewenttohisgrandfatherandtoldhimhisstoryangrily.“Heisreallybad,”... - ______[答案] 有一天,一个男孩与他的同学有了争吵,然后他去找他的外公,很生气地告诉了他这件事情,“他真的很坏”这个男孩说“还有,我恨他”. 外公说“让我告诉你个故事.当我小的时候,我也是,有时会因为别人做的一些事而...

(乌哈习14729098500)《论文》翻译 - ______ 首先给你纠正一下...这叫论语 不是论文... 子曰:“学而时习之,不亦说乎?有朋自远方来,不亦乐乎?人不知而不悦,不亦君子乎?” 孔子说:经常学习,不是愉悦的事吗?(说通悦)有远方的朋友来拜访,不是快乐得事吗?别人不了解我,但...

(乌哈习14729098500)翻译一段话,论文摘要,高分悬赏 ______ this is only in simple english,some spelling might not accurate,pls do not use it as accademy project...if you dont mind..but i dont really agree with the starting phrase,as decline of traditional retail business was mainly due to the direct competition of ...

(乌哈习14729098500)请翻译下以下论文谢辞,悬赏100分..网上在线翻译的不用了谢谢! ______ The completion of the paper is due to the information and knowledge of the professors in the Suzhou Agricultural Carrer Technique College. Under their help, I have accumulated all the demanding knowledge for his paper. I also benefited from my ...