Semi-Supervised Learning and Active Learning

Slides:



Advertisements
Similar presentations
Pat Langley Computational Learning Laboratory Center for the Study of Language and Information Stanford University, Stanford, California
Advertisements

Guiding Semi- Supervision with Constraint-Driven Learning Ming-Wei Chang,Lev Ratinow, Dan Roth.
On-line learning and Boosting
also known as the “Perceptron”
A Geometric Perspective on Machine Learning 何晓飞 浙江大学计算机学院 1.
PAC Learning adapted from Tom M.Mitchell Carnegie Mellon University.
Paper presentation for CSI5388 PENGCHENG XI Mar. 23, 2005
Introduction to Boosting Slides Adapted from Che Wanxiang( 车 万翔 ) at HIT, and Robin Dhamankar of Many thanks!
Multiple Criteria for Evaluating Land Cover Classification Algorithms Summary of a paper by R.S. DeFries and Jonathan Cheung-Wai Chan April, 2000 Remote.
Semi-supervised learning and self-training LING 572 Fei Xia 02/14/06.
Semi Supervised Learning Qiang Yang –Adapted from… Thanks –Zhi-Hua Zhou – ople/zhouzh/ –LAMDA.
Active Learning with Support Vector Machines
Data Mining with Decision Trees Lutz Hamel Dept. of Computer Science and Statistics University of Rhode Island.
Adaboost and its application
Three kinds of learning
Semi-Supervised Clustering Jieping Ye Department of Computer Science and Engineering Arizona State University
Introduction to Boosting Aristotelis Tsirigos SCLT seminar - NYU Computer Science.
Optimal Adaptation for Statistical Classifiers Xiao Li.
For Better Accuracy Eick: Ensemble Learning
Machine Learning CS 165B Spring 2012
Transfer Learning From Multiple Source Domains via Consensus Regularization Ping Luo, Fuzhen Zhuang, Hui Xiong, Yuhong Xiong, Qing He.
Active Learning for Class Imbalance Problem
Machine Learning1 Machine Learning: Summary Greg Grudic CSCI-4830.
Smart RSS Aggregator A text classification problem Alban Scholer & Markus Kirsten 2005.
Combined Active and Semi-Supervised Learning using Particle Walking Temporal Dynamics Fabricio Department of Statistics, Applied.
Use of web scraping and text mining techniques in the Istat survey on “Information and Communication Technology in enterprises” Giulio Barcaroli(*), Alessandra.
NEURAL NETWORKS FOR DATA MINING
Page 1 Ming Ji Department of Computer Science University of Illinois at Urbana-Champaign.
Active Learning An example From Xu et al., “Training SpamAssassin with Active Semi- Supervised Learning”
Benk Erika Kelemen Zsolt
Machine Learning.
 2003, G.Tecuci, Learning Agents Laboratory 1 Learning Agents Laboratory Computer Science Department George Mason University Prof. Gheorghe Tecuci 5.
1 KDD-09, Paris France Quantification and Semi-Supervised Classification Methods for Handling Changes in Class Distribution Jack Chongjie Xue † Gary M.
1 COMP3503 Semi-Supervised Learning COMP3503 Semi-Supervised Learning Daniel L. Silver.
Ensemble Learning Spring 2009 Ben-Gurion University of the Negev.
Particle Competition and Cooperation in Networks for Semi-Supervised Learning with Concept Drift Fabricio Breve 1,2 Liang Zhao 2
ISQS 6347, Data & Text Mining1 Ensemble Methods. ISQS 6347, Data & Text Mining 2 Ensemble Methods Construct a set of classifiers from the training data.
SemiBoost : Boosting for Semi-supervised Learning Pavan Kumar Mallapragada, Student Member, IEEE, Rong Jin, Member, IEEE, Anil K. Jain, Fellow, IEEE, and.
HAITHAM BOU AMMAR MAASTRICHT UNIVERSITY Transfer for Supervised Learning Tasks.
COP5992 – DATA MINING TERM PROJECT RANDOM SUBSPACE METHOD + CO-TRAINING by SELIM KALAYCI.
Ensemble Methods.  “No free lunch theorem” Wolpert and Macready 1995.
1 Unsupervised Learning and Clustering Shyh-Kang Jeng Department of Electrical Engineering/ Graduate Institute of Communication/ Graduate Institute of.
Agnostic Active Learning Maria-Florina Balcan*, Alina Beygelzimer**, John Langford*** * : Carnegie Mellon University, ** : IBM T.J. Watson Research Center,
HMMs and SVMs for Secondary Structure Prediction
Classification using Co-Training
Web-Mining Agents: Transfer Learning TrAdaBoost R. Möller Institute of Information Systems University of Lübeck.
SemiBoost : Boosting for Semi-supervised Learning Pavan Kumar Mallapragada, Student Member, IEEE, Rong Jin, Member, IEEE, Anil K. Jain, Fellow, IEEE, and.
Adaboost (Adaptive boosting) Jo Yeong-Jun Schapire, Robert E., and Yoram Singer. "Improved boosting algorithms using confidence- rated predictions."
Neural network based hybrid computing model for wind speed prediction K. Gnana Sheela, S.N. Deepa Neurocomputing Volume 122, 25 December 2013, Pages 425–429.
Introduction to Machine Learning, its potential usage in network area,
Data Mining Practical Machine Learning Tools and Techniques
CS 9633 Machine Learning Support Vector Machines
Machine Learning with Spark MLlib
Semi-Supervised Clustering
Research on Machine Learning and Deep Learning
Sofus A. Macskassy Fetch Technologies
Advanced data mining with TagHelper and Weka
Semi-supervised Machine Learning Gergana Lazarova
Combining Labeled and Unlabeled Data with Co-Training
SSL Chapter 4 Risk of Semi-supervised Learning: How Unlabeled Data Can Degrade Performance of Generative Classifiers.
Source: Procedia Computer Science(2015)70:
Active Learning Intrusion Detection using k-Means Clustering Selection
Introductory Seminar on Research: Fall 2017
Data Mining Practical Machine Learning Tools and Techniques
INTRODUCTION.
Ensemble learning.
How To Extend the Training Data
Concave Minimization for Support Vector Machine Classifiers
Artificial Intelligence 10. Neural Networks
Semi-Supervised Learning
Presentation transcript:

Semi-Supervised Learning and Active Learning 2007/3/20 Semi-Supervised Learning and Active Learning Machine Learning Center Faculty of Mathematics and Computer Science Hebei University Bin Wu

Labeling the examples in the training set L 2007/3/20 Labeling the examples in the training set L Time consuming Tedious Cost consuming Error prone process learn a target concept based on as few as possible labeled examples

semi-supervised learning 2007/3/20 reduce the need for labeled training data semi-supervised learning  active learning

semi-supervised learning 2007/3/20 semi-supervised learning

Semi-supervised Learning 2007/3/20 Labeled data set Unlabeled data set L h Select n (xi, h(xi)) n samples Labeling

Semi-supervised Learning 2007/3/20 Semi-supervised Learning Loop for N iterations Let h be the classifier obtained by training L on L Let MCP be the n examples in U on which h makes the most confident predictions For each x∈MCP do Remove x from U Add <x, h(x)> to L

2007/3/20

Active Learning L h Labeled data set Unlabeled data set Select 2007/3/20 Labeled data set Unlabeled data set L h Select (xi, c(xi)) 1 samples Labeling

Active Learning LOOP for N iterations Selective Sampling 2007/3/20 Active Learning Selective Sampling LOOP for N iterations let h be the classifier obtained by training L on L let remove q from U and ask the user for its label c(q) add <q, c(q)> to L

Common characteristics 2007/3/20 Common characteristics use an additional unlabeled set U try to maximize the accuracy

boosts the accuracy of a supervised 2007/3/20 Semi-supervised Learning boosts the accuracy of a supervised learner based on an additional set of unlabeled examples Active Learning minimizes the amount of labeled data by asking the user to label only the most informative examples

2007/3/20 Reference [1] Learning with labeled and unlabeled data, Matthias Seeger, Institute for Adaptive and Neural ComputationUniversity of Edinburgh, December 19, 2002 [2] Xiaojin Zhu, Semi-Supervised Learning Literature Survey, Computer Sciences, University of Wisconsin-Madiso, 2005, http://www.cs.wisc.edu/$\sim$jerryzhu/pub/ssl\_survey.pdf [3] M. Hasenjäger, Helge Ritter, Active Learning in Neural Networks, http://www.techfak.uni-bielefeld.de/ags/ni/projects/actlearn/ [4] M. Hasenjäger, H. Ritter, Active Learning with Local Models, Neural Processing Letters, 7, 107-117, 1998 [5] Ion Alexandru Muslea, ACTIVE LEARNING WITH MULTIPLE VIEWS, the Degree DOCTOR OF PHILOSOPHY, FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA, December 2002