Transfer Learning. keywords: transfer learning, inductive transfer, domain adaptation, multitask learning, semi-supervised learning

Σχετικά έγγραφα
IF(Ingerchange Format) [7] IF C-STAR(Consortium for speech translation advanced research ) [8] IF 2 IF

Discriminative Language Modeling Based on Risk Minimization Training

ΠΤΥΧΙΑΚΗ/ ΙΠΛΩΜΑΤΙΚΗ ΕΡΓΑΣΙΑ

5 Haar, R. Haar,. Antonads 994, Dogaru & Carn Kerkyacharan & Pcard 996. : Haar. Haar, y r x f rt xβ r + ε r x β r + mr k β r k ψ kx + ε r x, r,.. x [,

Kernel orthogonal and uncorrelated neighborhood preservation discriminant embedding algorithm

A Sequential Experimental Design based on Bayesian Statistics for Online Automatic Tuning. Reiji SUDA,

Vol. 34 ( 2014 ) No. 4. J. of Math. (PRC) : A : (2014) Frank-Wolfe [7],. Frank-Wolfe, ( ).

2016 IEEE/ACM International Conference on Mobile Software Engineering and Systems


Evaluation of Expressing Uncertain Causalities as Conditional Causal Possibilities

ER-Tree (Extended R*-Tree)

A Method for Determining Service Level of Road Network Based on Improved Capacity Model

MIDI [8] MIDI. [9] Hsu [1], [2] [10] Salamon [11] [5] Song [6] Sony, Minato, Tokyo , Japan a) b)

Power allocation under per-antenna power constraints in multiuser MIMO systems

Quantum annealing inversion and its implementation

J. of Math. (PRC) Banach, , X = N(T ) R(T + ), Y = R(T ) N(T + ). Vol. 37 ( 2017 ) No. 5

2002 Journal of Software /2002/13(08) Vol.13, No.8. , )

Optimization, PSO) DE [1, 2, 3, 4] PSO [5, 6, 7, 8, 9, 10, 11] (P)

Anomaly Detection with Neighborhood Preservation Principle

Generalized Linear Model [GLM]

{takasu, Conditional Random Field

IPSJ SIG Technical Report Vol.2014-CE-127 No /12/6 CS Activity 1,a) CS Computer Science Activity Activity Actvity Activity Dining Eight-He


Stabilization of stock price prediction by cross entropy optimization

(Υπογραϕή) (Υπογραϕή) (Υπογραϕή)

Quick algorithm f or computing core attribute

ΔΙΠΛΩΜΑΤΙΚΕΣ ΕΡΓΑΣΙΕΣ

A Method for Creating Shortcut Links by Considering Popularity of Contents in Structured P2P Networks

3: A convolution-pooling layer in PS-CNN 1: Partially Shared Deep Neural Network 2.2 Partially Shared Convolutional Neural Network 2: A hidden layer o



ΘΑΛΗΣ Πανεπιστήμιο Πειραιά Μεθοδολογικές προσεγγίσεις για τη μελέτη της ευστάθειας σε προβλήματα λήψης αποφάσεων με πολλαπλά κριτήρια

Research on model of early2warning of enterprise crisis based on entropy

Research on Economics and Management

Gradient Domain Metropolis Light Transport

CAPM. VaR Value at Risk. VaR. RAROC Risk-Adjusted Return on Capital

Ανάλυση ευαισθησίας σε αναδρομικό νευρωνικό δίκτυο εκπαιδευμένο για αναγνώριση συναισθήματος

q norm regularizing least-square-support-vector-machine linear classifier algorithm via iterative reweighted conjugate gradient

Schedulability Analysis Algorithm for Timing Constraint Workflow Models

An Automatic Modulation Classifier using a Frequency Discriminator for Intelligent Software Defined Radio

Vol. 31,No JOURNAL OF CHINA UNIVERSITY OF SCIENCE AND TECHNOLOGY Feb

Area Location and Recognition of Video Text Based on Depth Learning Method

Studies on the Binding Mechanism of Several Antibiotics and Human Serum Albumin

Novel Ensemble Analytic Discrete Framelet Expansion for Machinery Fault Diagnosis 1

Bayesian Discriminant Feature Selection


ΕΛΛΗΝΙΚΗ ΔΗΜΟΚΡΑΤΙΑ ΠΑΝΕΠΙΣΤΗΜΙΟ ΚΡΗΤΗΣ. Μηχανική Μάθηση. Ενότητα 10: Support Vector Machines. Ιωάννης Τσαμαρδίνος Τμήμα Επιστήμης Υπολογιστών

Proposal of Terminal Self Location Estimation Method to Consider Wireless Sensor Network Environment

A research on the influence of dummy activity on float in an AOA network and its amendments

Stochastic Finite Element Analysis for Composite Pressure Vessel

8.1 The Nature of Heteroskedasticity 8.2 Using the Least Squares Estimator 8.3 The Generalized Least Squares Estimator 8.

Nonparametric Bayesian T-Process Algorithm for Heterogeneous Gene Regulatory Network

Aerodynamic Design Optimization of Aeroengine Compressor Rotor

HOSVD. Higher Order Data Classification Method with Autocorrelation Matrix Correcting on HOSVD. Junichi MORIGAKI and Kaoru KATAYAMA

Robust Robot Monte Carlo Localization

Buried Markov Model Pairwise

ΚΑΤΗΓΟΡΙΟΠΟΙΗΣΗ ΜΕ ΜΗΧΑΝΕΣ ΔΙΑΝΥΣΜΑΤΩΝ ΥΠΟΣΤΗΡΙΞΗΣ

A Survey of Recent Clustering Methods for Data Mining (part 2)

Nondeterministic Finite Automaton Event Detection in Focusing Region. Sequence Analysis. Sequence Analysis. Feature Extraction. Feature Extraction

[1] DNA ATM [2] c 2013 Information Processing Society of Japan. Gait motion descriptors. Osaka University 2. Drexel University a)


Detection and Recognition of Traffic Signal Using Machine Learning

Αυτόματη κατηγοριοποίηση στρατηγικών επίλυσης προβλημάτων από μαθητές με χρήση Δικτύων Bayes.

ΑΝΑΠΤΥΞΗ ΕΝΟΣ ΕΚΠΑΙΔΕΥΣΙΜΟΥ ΑΝΙΧΝΕΥΤΗ ΟΡΙΩΝ ΦΡΑΣΕΩΝ (TEXT CHUNKER) ΓΙΑ ΤΑ ΝΕΑ ΕΛΛΗΝΙΚΑ


Μέθοδοι εκμάθησης ταξινομητών από θετικά παραδείγματα με αριθμητικά χαρακτηριστικά ΔΙΠΛΩΜΑΤΙΚΗ ΕΡΓΑΣΙΑ

: Monte Carlo EM 313, Louis (1982) EM, EM Newton-Raphson, /. EM, 2 Monte Carlo EM Newton-Raphson, Monte Carlo EM, Monte Carlo EM, /. 3, Monte Carlo EM

coupon effects Fisher Cohen, Kramer and Waugh Ordinary Least Squares OLS log

Probabilistic Approach to Robust Optimization

( ) , ) , ; kg 1) 80 % kg. Vol. 28,No. 1 Jan.,2006 RESOURCES SCIENCE : (2006) ,2 ,,,, ; ;

!" #$ : ( )

BCI On Feature Extraction from Multi-Channel Brain Waves Used for Brain Computer Interface

Gemini, FastMap, Applications. Εαρινό Εξάμηνο Τμήμα Μηχανικών Η/Υ και Πληροϕορικής Πολυτεχνική Σχολή, Πανεπιστήμιο Πατρών

Ταξινόμηση και διαχρονική παρακολούθηση των βοσκόμενων δασικών εκτάσεων στη λεκάνη απορροής του χειμάρρου Μπογδάνα Ν. Θεσσαλονίκης

A Non-Negative Sparse Neighbor Representation for Multi-Label Learning Algorithm

EL ECTR IC MACH IN ES AND CON TROL. System s vulnerability assessment of a ircraft guarantee system based on improved FPN

Motion analysis and simulation of a stratospheric airship

Maude 6. Maude [1] UIUC J. Meseguer. Maude. Maude SRI SRI. Maude. AC (Associative-Commutative) Maude. Maude Meseguer OBJ LTL SPIN


Customized Pricing Recommender System Simple Implementation and Preliminary Experiments

Πανεπιστήµιο Κρήτης - Τµήµα Επιστήµης Υπολογιστών. ΗΥ-570: Στατιστική Επεξεργασία Σήµατος. ιδάσκων : Α. Μουχτάρης. εύτερη Σειρά Ασκήσεων.

ΑΡΧΙΜΗ ΗΣ - ΕΝΙΣΧΥΣΗ ΕΡΕΥΝΗΤΙΚΩΝ ΟΜΑ ΩΝ ΣΤΑ ΤΕΙ. Υποέργο: «Ανάκτηση και προστασία πνευµατικών δικαιωµάτων σε δεδοµένα

Διαχείριση ενεργειακών πόρων & συστημάτων Πρακτικά συνεδρίου(isbn: )

ΠΟΛΥΤΕΧΝΕΙΟ ΚΡΗΤΗΣ ΣΧΟΛΗ ΜΗΧΑΝΙΚΩΝ ΠΑΡΑΓΩΓΗΣ ΚΑΙ ΔΙΟΙΚΗΣΗΣ

Concomitants of Dual Generalized Order Statistics from Bivariate Burr III Distribution

Generalized Fibonacci-Like Polynomial and its. Determinantal Identities

Περίληψη ιπλωµατικής Εργασίας

th International Conference on Machine Learning and Applications. E d. h. U h h b w k. b b f d h b f. h w k by v y

C3, I2, J2, J3 :JEL.

Toward a SPARQL Query Execution Mechanism using Dynamic Mapping Adaptation -A Preliminary Report- Takuya Adachi 1 Naoki Fukuta 2.

Security in the Cloud Era

α + α+ α! (=+9 [1] ι «Analyze-Regression-Linear». «Dependent» ι η η η!ηη ι «Independent(s)» η!ηη. # ι ι ι!η " ι ιηη, ι!" ι ηιι. 1 SPSS ι η η ι ιηη ι η

b,% SIR 2 MOTDPC (CDMA 6 ) Aein CDMA Journal of Nonlinear Systems in Elect. Eng., Vol. 1, No 2, Fall 2013

Discrete Fourier Transform { } ( ) sin( ) Discrete Sine Transformation. n, n= 0,1,2,, when the function is odd, f (x) = f ( x) L L L N N.

P rogresses in H azardou s M aterials L ogistics R esearch

ΓΙΑΝΝΟΥΛΑ Σ. ΦΛΩΡΟΥ Ι ΑΚΤΟΡΑΣ ΤΟΥ ΤΜΗΜΑΤΟΣ ΕΦΑΡΜΟΣΜΕΝΗΣ ΠΛΗΡΟΦΟΡΙΚΗΣ ΤΟΥ ΠΑΝΕΠΙΣΤΗΜΙΟΥ ΜΑΚΕ ΟΝΙΑΣ ΒΙΟΓΡΑΦΙΚΟ ΣΗΜΕΙΩΜΑ

Performance of Charcoal Cookstoves for Haiti, Part 1: Results from the Water Boiling Test

CSJ. Speaker clustering based on non-negative matrix factorization using i-vector-based speaker similarity

EQUIVALENT MODEL OF HVDC-VSC AND ITS HYBRID SIMULATION TECHNIQUE

Lecture Notes for Chapter 8

Transcript:

1 Transfer Learnng Toshhro Kamshma Natonal Insttute of Advanced Industral Scence and Technology (AIST) mal@kamshma.net, http://www.kamshma.net/ keywords: transfer learnng, nductve transfer, doman adaptaton, multtask learnng, sem-supervsed learnng 1. (transfer learnng) [TT 05] the problem of retanng and applyng the knowledge learned n one or more tasks to effcently develop an effectve hypothess for a new task CPU montor CPU montor CPU Web (sem-supervsed learnng) [Chapelle 06] [ 05, 97] 1995 NIPS [LtL 95] 10 (nductve transfer) (doman adaptaton) (multtask learnng) knowledge transfer, learnng to learn lfetme learnng (covarate shft) [Shmodara 00, 06] [Heckman 79, 09] Pan & Yang [Pan 08b] Daumé [Daumé]

2 25 4 2010 7 2 3 [Bshop 08, 03, 06, 09] 2. (a) (b) 1 2 1 (source doman) (target doman) (S) (T ) D {S,T } X (D) X (D) M Y (D) Y (D) {0,1} { 1,+1} (x (D),y (D) ) (X (D),Y (D) ) x (D) X (D) N (D) N 2 2 (multtask learnng) Daumé / Daumé Pan & Yang A B C B C A A C B A B C A B C A B C B C A A (B C) 1(a) A A B C 1(b) 100 101 (meta learnng) [Vlalta 02] No Free Lunch [Wolpert 97]

3 1 (1) S+T + (2) S+T (3) S T + (4) S T 2 3 4 Daumé 1 (1) S+T + {(x (S),y (S) )} {(x (T ),y (T ) )} Pan & Yang Daumé (1) (nductve transfer learnng) (2) (transductve transfer learnng) (3) (self-taught learnng) (4) (unsupervsed transfer learnng) (1) S+T + Pr (S) [X,Y ] Pr (T ) [X,Y ] X (S) X (T ) Y (S) Y (T ) (2) S+T Daumé Pan & Yang Pr (S) [Y X] = Pr (T ) [Y X] Pr (S) [X] Pr (T ) [X] X (3) S T + Daumé Pr (S) [X] = Pr (T ) [X] Rana [Rana 07] Pr (S) [X] = Pr (T ) [X] Y (S) Y (T ) Rana Pr (S) [X] Pr (T ) [X] Rana Pr (S) [X] Pr (T ) [X] (4) S T Pr (S) [X] = Pr (T ) [X] Daumé X (S) X (T ) [Wang 08] 2 4 [Caruana 97] (negatve transfer) [Rosensten 05] 2

4 25 4 2010 7 2 Pan & Yang Daumé (feature-based) (nstance-based) (separated) (ntegrated) 3(a) 3(b) (a) (b) 3 3. 3 1 [Pan 08b, Daumé, 09] [TT 05, TSL 09, TM 09] 90 [Caruana 96, Munro 97] Thrun [Thrun 96] 1 k explanaton-based Caruana [Caruana 97]

5 (1) (data amplfcaton) (2) (attrbute selecton) (3) (eavesdroppng) (4) (representaton bas) 4 3 2 [Daumé 07] 0 3 (x (T ),y (T ) ) ( x (T ),0,x (T ),y (T ) ) (x (S),y (S) ) ( x (S),x (S),0,y (S) ) 0 3 3 3 1 [Caruana 97] K Pr[x,y Θ ]Pr[Θ Ψ] Θ Ψ [Rana 06] [Daumé 06] f Gbbs exp(λ f) λ Gbbs [Xue 08] plsa [Hofmann 99] must/cannot [Wagstaff 01] [Ando 08] [Tshby 99] [Da 07a] (co-clusterng) [Argyrou 07] K k f k (x) = M m a mk(u x) M u a mk 0 u a mk U A K k=1 =1 N L(y k,a k (U x k )) + γ A 2 2,1 1 2 L 2 L 1 0 [Lng 08]

6 25 4 2010 7 normalzed cut [Sh 00] W W (S) D = dag(w1) D (S) = dag(w (S) 1), x x (D W)x x Dx + β U x + λ x (D (S) W (S) )x x D (S) x 1 Raylegh W 0 1 x 2 3 1 β λ 2 3 [Rana 07] m M b 1,...,b m mn a,b x (S) m a j b j 2 2 + β a 1, s.t. b j 2 1 j a j m j b arg mn x (T ) c j c j b j 2 2 + γ c 1 m c y (T ) [Ando 05] [Argyrou 08] [Satpal 07] [Wang 08] Fsher [Do 06, Pan 08a] [Rückert 08] 3 4 AdaBoost [Freund 96, 99] TrAdaBoost [Da 07b] TrAdaBoost T t h t (x ) {0,1} h t ( ) ϵ t < 1/2 β t = ϵ t /(1 ϵ t ) 1/β t 1/(1 + (2lnt)/T ) 1 0 T t= T/2 β h t(x) t T t= T/2 β 1/2 t RankBoost[Freund 03] [ 09] [Breman 96] TrBagg[Kamshma 09] TrBagg

7 [Eaton 08] [Gao 08] (covarate shft) [Shmodara 00] Pr (S) [X] Pr (T ) [X] Pr (S) [Y X] = Pr (T ) [Y X] θ N (T ) Pr (T ) [x ] Pr (S) [x ] loss(y(s),x (S) ;θ) [ 06] [Sugyama 07b, Huang 07, 07a] (sample selecton bas) [Heckman 79, 09] [Zadrozny 04] x y s {0,1} (x,y) s = 1 s = 0 x s y Pr[y s,x] = Pr[y x] Pr[y x] Pr[y x] Pr[x] Pr[y s,x] = Pr[y x] Pr[y x] s Pr[x] s SVM SVM [Xng 07] brdged refnement 3 5 Mgratory-Logt [Lao 05] µ w max w,µ σ(y (T ) w x (T ) ) + lnσ(y (S) w x (S) + y (S) µ ) 1 subject to y (S) N (S) µ C, C 0, y (S) µ 0 y (D) { 1,+1} y w x y (S) µ (S) x (S) y (S) w x (S) y (S) µ (S) N (S) C C [Wu 04] 3 3 [Wu 04] 3 6 (currculum learnng) [Bengo 09]

8 25 4 2010 7 3 3 [Rosensten 05] (negatve transfer) [Sh 08] [Bltzer 08, Crammer 08] [Ben-Davd 07, Da 07b] 4. [TSL 09, TM 09] [Ando 05] Ando, R. K. and Zhang, T.: A Framework for Learnng Predctve Structures from Multple Tasks and Unlabeled Data, Journal of Machne Learnng Research, Vol. 6, pp. 1817 1853 (2005) [Ando 08] Ando, S. and Suzuk, E.: Unsupervsed Cross-doman Learnng by Interacton Informaton Co-clusterng, n Proc. of The 8th IEEE Int l Conf. on Data Mnng, pp. 13 22 (2008) [Argyrou 07] Argyrou, A., Evgenou, T., and Pontl, M.: Mult-Task Feature Learnng, n Advances n Neural Informaton Processng Systems 19, pp. 41 48 (2007) [Argyrou 08] Argyrou, A., Maurer, A., and Pontl, M.: An Algorthm for Transfer Learnng n a Heterogeneous Envronment, n Proc. of The ECML/PKDD2008, Part I, pp. 71 85 (2008), [LNAI 5211] [ 03],,,, 6, (2003) [ 09] Doman Adaptaton, 2009, pp. 69 72 (2009) [Ben-Davd 07] Ben-Davd, S., Bltzer, J., Crammer, K., and Perera, F.: Analyss of Representatons for Doman Adaptaton, n Advances n Neural Informaton Processng Systems 19, pp. 137 144 (2007) [Bengo 09] Bengo, Y., Louradour, J., Collobert, R., and Weston, J.: Currculum Learnng, n Proc. of The 26th Int l Conf. on Machne Learnng, pp. 41 48 (2009) [Bshop 08] Bshop, C. M.:, (2007 2008), [Bltzer 08] Bltzer, J., Crammer, K., Kulesza, A., Perera, F., and Wortman, J.: Learnng Bounds for Doman Adaptaton, n Advances n Neural Informaton Processng Systems 20, pp. 129 136 (2008) [Breman 96] Breman, L.: Baggng Predctors, Machne Learnng, Vol. 24, pp. 123 140 (1996) [Caruana 96] Caruana, R., Baluja, S., and Mtchell, T.: Usng The Future to Sort Out The Present: Rankprop and Multtask Learnng for Medcal Rsk Evaluaton, n Advances n Neural Informaton Processng Systems 8, pp. 959 965 (1996) [Caruana 97] Caruana, R.: Multtask Learnng, Machne Learnng, Vol. 28, pp. 41 75 (1997) [Chapelle 06] Chapelle, O., Schölkopf, B., and Zen, A. eds.: Semsupervsed Learnng, MIT Press (2006) [Crammer 08] Crammer, K., Kearns, M., and Wortman, J.: Learnng from Multple Sources, Journal of Machne Learnng Research, Vol. 9, pp. 1757 1774 (2008) [Da 07a] Da, W., Xue, G.-R., Yang, Q., and Yu, Y.: Co-clusterng based Classfcaton for Out-of-doman Documents, n Proc. of The 13th Int l Conf. on Knowledge Dscovery and Data Mnng, pp. 210 219 (2007) [Da 07b] Da, W., Yang, Q., Xue, G.-R., and Yu, Y.: Boostng for Transfer Learnng, n Proc. of The 24th Int l Conf. on Machne Learnng, pp. 193 200 (2007) [Daumé] Daumé, H., III: natural language processng blog, http://nlpers.blogspot.com/search/label/ doman%20adaptaton [Daumé 06] Daumé, H., III and Marcu, D.: Doman Adaptaton for Statstcal Classfers, Journal of Artfcal Intellgence Research, Vol. 26, pp. 101 126 (2006) [Daumé 07] Daumé, H., III: Frustratngly Easy Doman Adaptaton, n Proc. of the 45th Annual Meetng of the Assocaton of Computatonal Lngustcs, pp. 256 263 (2007) [Do 06] Do, C. B. and Ng, A. Y.: Transfer Learnng for Text Classfcaton, n Advances n Neural Informaton Processng Systems 18, pp. 299 306 (2006) [Eaton 08] Eaton, E., desjardns, M., and Lane, T.: Modelng Transfer Relatonshps Between Learnng Tasks for Improved Inductve Transfer, n Proc. of The ECML/PKDD2008, Part I, pp. 317 332 (2008), [LNAI 5211] [Freund 96] Freund, Y. and Schapre, R. E.: Experments wth a New Boostng Algorthm, n Proc. of The 13th Int l Conf. on Machne Learnng, pp. 148 156 (1996) [ 99] Y., R.,,, Vol. 14, No. 5, pp. 771 780 (1999) [Freund 03] Freund, Y., Iyer, R., Schapre, R. E., and Snger, Y.: An

9 Effcent Boostng Algorthm for Combnng Preferences, Journal of Machne Learnng Research, Vol. 4, pp. 933 969 (2003) [Gao 08] Gao, J., Fan, W., Jang, J., and Han, J.: Knowledge Transfer va Multple Model Local Structure Mappng, n Proc. of The 14th Int l Conf. on Knowledge Dscovery and Data Mnng, pp. 283 291 (2008) [Heckman 79] Heckman, J.: Sample Selecton Bas as a Specfcaton Error, Econometrca, Vol. 47, pp. 153 161 (1979) [Hofmann 99] Hofmann, T.: Probablstc Latent Semantc Analyss, n Uncertanty n Artfcal Intellgence 15, pp. 289 296 (1999) [ 09],, (2009) [Huang 07] Huang, J., Smola, A. J., Gretton, A., Borgwardt, K. M., and Schölkopf, B.: Correctng Sample Selecton Bas by Unlabeled Data, n Advances n Neural Informaton Processng Systems 19, pp. 601 608 (2007) [Kamshma 09] Kamshma, T., Hamasak, M., and Akaho, S.: TrBagg: A Smple Transfer Learnng Method and Its Applcaton to Personalzaton n Collaboratve Taggng, n Proc. of The 9th IEEE Int l Conf. on Data Mnng, pp. 219 228 (2009) [ 09],,, R, 5, (2009) [Lao 05] Lao, X., Xue, Y., and Carn, L.: Logstc Regresson wth an Auxlary Data Source, n Proc. of The 22nd Int l Conf. on Machne Learnng, pp. 505 512 (2005) [Lng 08] Lng, X., Da, W., Xue, G.-R., Yang, Q., and Yu, Y.: Spectral Doman-Transfer Learnng, n Proc. of The 14th Int l Conf. on Knowledge Dscovery and Data Mnng, pp. 488 496 (2008) [LtL 95] Learnng to Learn: Knowledge Consoldaton and Transfer n Inductve Systems, http://socrates.acadau. ca/courses/comp/dslver/nips95 LTL/transfer. workshop.1995.html (1995) [ 06],,,, (2006) [Munro 97] Munro, P. W. and Parmanto, B.: Competton Among Networks Improves Commttee Performance, n Advances n Neural Informaton Processng Systems 9, pp. 592 598 (1997) [ 97],, Vol. 38, No. 7, pp. 557 588 (1997) [Pan 08a] Pan, S. J., Kwok, J. T., and Yang, Q.: Transfer Learnng va Dmensonalty Reducton, n Proc. of the 23rd Natonal Conf. on Artfcal Intellgence, pp. 677 682 (2008) [Pan 08b] Pan, S. J. and Yang, Q.: A Survey on Transfer Learnng, Techncal Report HKUST-CS08-08, Dept. of Computer Scence and Engneerng, Hong Kong Unv. of Scence and Technology (2008) [Rana 06] Rana, R., Ng, A. Y., and Koller, D.: Constructng Informatve Prors usng Transfer Learnng, n Proc. of The 23rd Int l Conf. on Machne Learnng, pp. 713 720 (2006) [Rana 07] Rana, R., Battle, A., Lee, H., Packer, B., and Ng, A. Y.: Self-taught Learnng: Transfer Learnng from Unlabeled Data, n Proc. of The 24th Int l Conf. on Machne Learnng, pp. 759 766 (2007) [Rosensten 05] Rosensten, M. T., Marx, Z., Kaelblng, L. P., and Detterch, T. G.: To Transfer or Not To Transfer, n NIPS-2005 Workshop on Inductve Transfer: 10 Years Later (2005) [Rückert 08] Rückert, U. and Kramer, S.: Kernel-Based Inductve Transfer, n Proc. of The ECML/PKDD2008, Part II, pp. 220 233 (2008), [LNAI 5212] [Satpal 07] Satpal, S. and Sarawag, S.: Doman Adaptaton of Condtonal Probablty Models va Feature Subsettng, n Proc. of the 11th European Conf. on Prncples of Data Mnng and Knowledge Dscovery, pp. 224 235 (2007), [LNAI 4702] [Sh 00] Sh, J. and Malk, J.: Normalzed Cuts and Image Segmentaton, IEEE Trans. on Pattern Analyss and Machne Intellgence, Vol. 22, No. 8, pp. 888 905 (2000) [Sh 08] Sh, X., Fan, W., and Ren, J.: Actvely Transfer Doman Knowledge, n Proc. of The ECML/PKDD2008, Part II, pp. 342 357 (2008), [LNAI 5212] [Shmodara 00] Shmodara, H.: Improvng Predctve Inference under Covarate Shft by Weghtng the Log-Lkelhood Functon, J. of Statstcal Plannng and Inference, Vol. 90, pp. 227 244 (2000) [ 06],, Vol. 13, No. 3, pp. 111 118 (2006) [ 07a],, Vol. 18, No. 10, pp. 1 6 (2007) [Sugyama 07b] Sugyama, M., Krauledat, M., and Müller, K. R.: Covarate Shft Adaptaton by Importance Weghted Cross Valdaton, Journal of Machne Learnng Research, Vol. 8, pp. 985 1005 (2007) [ 09],,,, n WdbDB Forum 2009 (2009) [Thrun 96] Thrun, S.: Is Learnng The n-th Thng Any Easer Than Learnng The Frst?, n Advances n Neural Informaton Processng Systems 8, pp. 640 646 (1996) [Tshby 99] Tshby, N., Perera, F. C., and Balek, W.: The Informaton Bottleneck Method, n Proc. of The 37th Annual Allerton Conference on Communcatons, Control and Computng (1999) [TM 09] ICDM 2009 Workshop: Int l Workshop on Transfer Mnng, http://www.cse.ust.hk/ snnopan/cfp/ cdm09wtm.html (2009) [TSL 09] NIPS 2009 Workshop: Transfer Learnng for Structured Data, http://www.cse.ust.hk/ snnopan/ nps09tlsd/ (2009) [TT 05] NIPS 2005 Workshop Inductve Transfer: 10 Years Later, http://trl.acadau.ca/tws05/ (2005) [Vlalta 02] Vlalta, R. and Drss, Y.: A Perspectve Vew and Survey of Meta-Learnng, Artfcal Intellgence Revew, Vol. 18, pp. 77 95 (2002) [Wagstaff 01] Wagstaff, K., Carde, C., Rogers, S., and Schroedl, S.: Constraned K-means Clusterng wth Background Knowledge, n Proc. of The 18th Int l Conf. on Machne Learnng, pp. 577 584 (2001) [Wang 08] Wang, Z., Song, Y., and Zhang, C.: Transferred Dmensonalty Reducton, n Proc. of The ECML/PKDD2008, Part II, pp. 550 565 (2008), [LNAI 5212] [ 05],,,,,,, (2005) [Wolpert 97] Wolpert, D. H. and Macready, W. G.: No Free Lunch Theorems for Optmzaton, IEEE Transactons on Evolutonary Computaton, Vol. 1, pp. 67 82 (1997) [Wu 04] Wu, P. and Detterch, T. G.: Improvng SVM Accuracy by Tranng on Auxlary Data Sources, n Proc. of The 21st Int l Conf. on Machne Learnng, pp. 871 878 (2004) [Xng 07] Xng, D., Da, W., Xue, G.-R., and Yu, Y.: Brdged Refnement for Transfer Learnng, n Proc. of the 11th European Conf. on Prncples of Data Mnng and Knowledge Dscovery, pp. 324 335 (2007), [LNAI 4702] [Xue 08] Xue, G.-R., Da, W., Yang, Q., and Yu, Y.: Topc-brdged PLSA for Cross-Doman Text Classfcaton, n Proc. of The 31th Annual ACM SIGIR Conf. on Research and Development n Informaton Retreval, pp. 627 634 (2008) [Zadrozny 04] Zadrozny, B.: Learnng and Evaluatng Classfers under Sample Selecton Bas, n Proc. of The 21st Int l Conf. on Machne Learnng, pp. 903 910 (2004) 1968 1992 1994 2001 ( ) AAAI, ACM