semi-supervised-learning半监督学习.pptxVIP

Semi-supervised Learning2017.11;Traditional machine learning paradigms;Big data era, obtaining data is getting easier and easier.;Labeling the data is difficult! expensive, time consuming, sometimes need experts. For example, medical image analysis, webpage recommendation.; ;Outline;Concept of semi-supervised learning;Usefulness of unlabeled data;Outline;Semi-supervised learning methods;Semi-supervised learning methods;Disagreement-based methods;;;;;Typical algorithm：Co-training;An example of multi-view data;web pages driverless vehicle; Co-training trains two classifiers separately on two views, then uses the predictions of each classifier on unlabeled examples to augment the training set of the other.;互帮互助 共同进步;;;;Tri-training;Tri-training;;TriTrain? TriTrain is a semi-supervised algorithm, which iteratively refines each of the three component classifiers , and finally combines their prediction via majority voting. CoForest CoForest is a semi-supervised algorithm, which exploits the power of ensemble learning and large amount of unlabeled data available to produce hypothesis with better performance. COREG COREG is a co-training style semi-supervised regression algorithm, which employs two k-NN regressors using different distance metrics to select the most confidently labeled unlabeled examples for each other.;Semi-supervised learning methods;Semi-supervised SVM;TSVM algorithm ;+;TSVM algorithm ;Semi-supervised learning methods;Generative methods;;Basis knowledge:iid;;Gaussian Model;;;;;Advantages: Simple and easy, better than other methods especially when labeled data is very few. Shortcomings: Model assumption must be right , if not, performance will be reduced by using unlabeled data.;Dataset can be mapped to a graph, data is presented by a point, the similarity between two data ( or distance) can present an edge. ;;; Label propagation algorithm (LPA) Intuitively, similar samples should have similar labels. From labe