Introduction to Transfer Learning (Part 2) For 2012 Dragon .ppt

《Introduction to Transfer Learning (Part 2) For 2012 Dragon .ppt》由会员分享，可在线阅读，更多相关《Introduction to Transfer Learning (Part 2) For 2012 Dragon .ppt（27页珍藏版）》请在麦多课文档分享上搜索。

1、1,Introduction to Transfer Learning (Part 2) For 2012 Dragon Star Lectures,Qiang YangHong Kong University of Science and Technology Hong Kong, China http:/www.cse.ust.hk/qyang,Domain Adaptation in NLP,Applications,Automatic Content Extraction Sentiment Classification Part-Of-Speech Tagging NER Quest

2、ion Answering Classification Clustering,Selected Methods,Domain adaptation for statistical classifiers Hal Daume III & Daniel Marcu, JAIR 2006, Jiang and Zhai, ACL 2007 Structural Correspondence Learning John Blitzer et al. ACL 2007 Ando and Zhang, JMLR 2005 Latent subspace Sinno Jialin Pan et al. A

3、AAI 08,2,Instance-transfer Approaches Wu and Dietterich ICML-04 J.Jiang and C. Zhai, ACL 2007 Dai, Yang et al. ICML-07,Differentiate the cost for misclassification of the target and source data,3,Uniform weights,Correct the decision boundary by re-weighting,Loss function on the target domain data,Lo

4、ss function on the source domain data,Regularization term,Cross-domain POS tagging,entity type classification Personalized spam filtering,TrAdaBoost Dai, Yang et al. ICML-07,4,Misclassified examples: increase the weights of the misclassified target data decrease the weights of the misclassified sour

5、ce data,Evaluation with 20NG: 22%8% http:/people.csail.mit.edu/jrennie/20Newsgroups/,Locally Weighted Ensemble Jing Gao, Wei Fan, Jing Jiang, Jiawei Han: Knowledge transfer via multiple model local structure mapping. KDD 2008,Graph-based weights approximationWeight of a model is proportional to the

6、similarity between its neighborhood graph and the clustering structure around x.,Transductive Transfer Learning Instance-transfer Approaches Sample Selection Bias / Covariance Shift Zadrozny ICML-04, Schwaighofer JSPI-00,Input: A lot of labeled data in the source domain and no labeled data in the ta

7、rget domain.Output: Models for use in the target domain data.Assumption: The source domain and target domain are the same. In addition, and are the same while and may be different caused by different sampling process (training data and test data).Main Idea: Re-weighting (important sampling) the sour

8、ce domain data.,Sample Selection Bias/Covariance Shift,To correct sample selection bias:How to estimate ? One straightforward solution is to estimate and , respectively. However, estimating density function is a hard problem.,weights for source domain data,Sample Selection Bias/Covariance Shift Kern

9、el Mean Match (KMM) Huang et al. NIPS 2006,Main Idea: KMM tries to estimate directly instead of estimating density function. It can be proved that can be estimated by solving the following quadratic programming (QP) optimization problem.Theoretical Support: Maximum Mean Discrepancy (MMD) Borgwardt e

10、t al. BIOINFOMATICS-06. The distance of distributions can be measured by Euclid distance of their mean vectors in a RKHS.,To match means between training and test data in a RKHS,9,Feature Space: Document-word co-occurrence,D_S,D_T,Knowledge transfer,Source,Target,10,10,Co-Clustering based Classifica

11、tion (KDD 2007),Co-clustering is applied between features (words) and target-domain documents Word clustering is constrained by the labels of in-domain (Old) documents The word clustering part in both domains serve as a bridge,Structural Correspondence Learning Blitzer et al. ACL 2007,SCL: Ando and

12、Zhang, JMLR 2005 Define pivot features: common in two domains Build Latent Space built from Pivot Features, and do mapping Build classifiers through the non-pivot Features,11,SCL Blitzer et al. EMNLP-06, Blitzer et al. ACL-07, Ando and Zhang JMLR-05,a) Heuristically choose m pivot features, which is

13、 task specific. b) Transform each vector of pivot feature to a vector of binary values and then create corresponding prediction problem.,Learn parameters of each prediction problem,Do Eigen Decomposition on the matrix of parameters and learn the linear mapping function.,Use the learnt mapping functi

14、on to construct new features and train classifiers onto the new representations.,Courtesy of Sinno Pan,Self-Taught Learning Feature-representation-transfer Approaches Unsupervised Feature Construction Raina et al. ICML-07,Three steps: Applying sparse coding Lee et al. NIPS-07 algorithm to learn high

15、er-level representation from unlabeled data in the source domain.Transforming the target data to new representations by new bases learnt in the first step. Traditional discriminative models can be applied on new representations of the target data with corresponding labels.,Courtesy of Sinno Pan,Step

16、1:Input: Source domain data and coefficient Output: New representations of the source domain data and new bases Step2:Input: Target domain data , coefficient and bases Output: New representations of the target domain data,Unsupervised Feature Construction Raina et al. ICML-07,Courtesy of Sinno Pan,“

17、Self-taught Learning” Raina et al. Self-Taught Learning ICML-07,Self-taught Learning: Courtesy of Raina,15,16,Examples of Higher Level Features Learned,Natural images.,Learnt bases: “Edges”,Handwritten characters.,Learnt bases: “Strokes”,Self-taught Learning: Courtesy of Raina,17,Latent Feature Spac

18、e TL Methods: Temporal Domain Distribution Changes,The mapping function f learned in the offline phase can be out of date. Recollecting the WiFi data is very expensive. How to adapt the model ?,Time,Night time period,Day time period,18,Transfer Component Analysis: Sinno Pan et al., IEEE Trans. NN 20

19、11,Source Domain data,Target Domain data,Observations,Latent factors,If two domains are related, ,Common latent factors across domains,Sinno Jialin Pan,19,Motivation (cont.),Source domain data,Target domain data,Observations,Latent factors,Some latent factors may preserve important properties (such

20、as variance, local topological structure) of the original data, while others may not.,Sinno Jialin Pan,PCA: Only Maximizing the Data Variance,20,Principal Component Analysis (PCA) Jolliffe. 02 aims to find a low-dimensional latent space where the variance of the projected data is maximized. Con: it

21、may not reduce the difference between domains.,Sinno Jialin Pan,21,Learning the Transform Mapping,How to estimate distance between distributions in the latent space?,How to solve the resultant optimization problem?,High level optimization problem,Sinno Jialin Pan,22,Semi-Supervised TCA,High level ob

22、jectives:,To measure label dependence using Hilbert-Schmidt Independence Criterion (HSIC),To measure the distance between domains using MMD,Sinno Jialin Pan,Blitzer, et al. Learning Bounds for Domain Adaptation. NIPS 2007,m=number of examples d(u_S, u_T) = domain distance 1-d=confidence e=error,23,I

23、nductive Transfer Learning Model-transfer Approaches Regularization-based Method Evgeiou and Pontil, KDD-04,Assumption: If t tasks are related to each other, then they may share some parameters among individual models. Assume be a hyper-plane for task , where and Encode them into SVMs:,Common part,S

24、pecific part for individual task,Regularization terms for multiple tasks,Inductive Transfer Learning Structural-transfer Approaches TAMAR Mihalkova et al. AAAI-07,Assumption: If the target domain and source domain are related, then there may be some relationship between domains that are similar, whi

25、ch can be used for transfer learningInput: Relational data in the source domain and a statistical relational model, Markov Logic Network (MLN), which has been learnt in the source domain. Relational data in the target domain.Output: A new statistical relational model, MLN, in the target domain.Goal:

26、 To learn a MLN in the target domain more efficiently and effectively.,TAMAR Mihalkova et al. AAAI-07,Two Stages: Predicate Mapping Establish the mapping between predicates in the source and target domain. Once a mapping is established, clauses from the source domain can be translated into the targe

27、t domain. Revising the Mapped Structure The clause-mapping from the source domain directly may not be completely accurate and may need to be revised, augmented , and re-weighted in order to properly model the target data.,TAMAR Mihalkova et al. AAAI-07,Source domain (academic domain),Target domain (movie domain),Mapping,Revising,