遥感文件模式识别boosting.pptxVIP

Boosting算法; 1. Boosting概述 2. Boosting应用研究 ;1. Boosting概述;1.1 源起;Bootstrap Estimation;; 　 Bagging方法的主要过程 　 i) 重复地从一个样本集合中，抽样n*??N 个样本，针对抽样的集合训练分类器Ci 　ii) 分类器进行等权投票;;Schapire 1989;1.1 源起;1.1 源起;为此，我们要弱化对学习器的要求：1. 我们不要求学习器输出零错误率的假设，只要求错误率被限制在某常数ε范围内，ε可为任意小。2. 不要求学习器对所有任意抽取的数据都能成功预测，只要求其失败的概率被限定在某个常数μ的范围内，μ可取任意小。 简而言之，我们只要求学习器可能学习到一个近似正确的假设，故得到了“可能近似正确学习”或PAC（ Probably Approximately Correct）学习。 一个可PAC学习的学习器要满足两个条件： 1. 学习器必须以任意高的概率输出一个错误率任意低的假设 2. 学习过程的时间最多以多项式方式增长 ;1.1 源起;1.1 概述;1.1 概述;1.1 概述;1.1 概述;;1.1 概述;1.1 概述;1.1 概述;;Adaboost Terms;AdaBoost Specifics - How does AdaBoost weight training examples optimally? ? Focus on difficult data points. The data points that have been misclassified most by the previous weak classifier. -How does AdaBoost combine these weak classifiers into a comprehensive prediction? ? Use an optimally weighted majority vote of weak classifier. ;1.1 概述;1.1 概述;1.1 概述;1.1 概述;1.1 概述;Adaboost - Adaptive Boosting;Key Questions Defining and Analyzing Boosting;Discrete Adaboost (DiscreteAB)(Friedman’s wording);Discrete Adaboost (DiscreteAB)(Freund and Schapire’s wording);Adaboost with Confidence Weighted Predictions (RealAB);Adaboost Variants Proposed By Friedman;Adaboost Variants Proposed By Friedman;Adaboost Variants Proposed By Friedman;2. Boosting应用研究;Viola-Jones Face Detection Algorithm;Viola Jones Technique Overview;Features;Integral Images;Fast Computation of Pixel Sums;Feature Extraction;Useful Features Learned by Boosting;Weak Learners for Face Detection;Reduction in Error as Boosting adds Classifiers;A Cascade of Classifiers;Training;Small Set of 111 Training Images;Sample Results Using the Viola-Jones Detector;More Detection Examples;Practical Implementation;?;近期研究应用;Original Papers and Surveys Y Freund, RE?Schapire, A decision-theoretic generalization of on-line learning and an application to boosting,?EuroCOLT, 1995 Y Freund, RE?Schapire,?Experiments with a new boosting algorithm, ICML, 1996 Robert E.?Schapire,?The bo