## William W. Cohen, Andrew McCallum, Sam T. Roweis

*Proceedings of the 25th International Conference on Machine Learning*

ICML, 2008.

@proceedings{ICML-2008, address = "Helsinki, Finland", editor = "William W. Cohen and Andrew McCallum and Sam T. Roweis", isbn = "978-1-60558-205-4", publisher = "{ACM}", series = "{ACM International Conference Proceeding Series}", title = "{Proceedings of the 25th International Conference on Machine Learning}", volume = 307, year = 2008, }

### Contents (157 items)

- ICML-2008-AdamsS #modelling #parametricity #process
- Gaussian process product models for nonparametric nonstationarity (RPA, OS), pp. 1–8.
- ICML-2008-AllauzenMT #kernel #predict #sequence
- Sequence kernels for predicting protein essentiality (CA, MM, AT), pp. 9–16.
- ICML-2008-AnWSWCD #analysis #image #kernel #multi #process
- Hierarchical kernel stick-breaking process for multi-task image analysis (QA, CW, IS, EW, LC, DBD), pp. 17–24.
- ICML-2008-Bach #graph #kernel
- Graph kernels between point clouds (FRB), pp. 25–32.
- ICML-2008-Bach08a #consistency #estimation #named
- Bolasso: model consistent Lasso estimation through the bootstrap (FRB), pp. 33–40.
- ICML-2008-BarrettN #learning #multi #policy
- Learning all optimal policies with multiple criteria (LB, SN), pp. 41–47.
- ICML-2008-BergeronZBB #multi #ranking
- Multiple instance ranking (CB, JZ, CMB, KPB), pp. 48–55.
- ICML-2008-BickelBLS #learning #multi
- Multi-task learning for HIV therapy screening (SB, JB, TL, TS), pp. 56–63.
- ICML-2008-BiggsGV #matrix
- Nonnegative matrix factorization via rank-one downdate (MB, AG, SAV), pp. 64–71.
- ICML-2008-BowlingJBS #evaluation #game studies
- Strategy evaluation in extensive games with importance sampling (MHB, MJ, NB, DS), pp. 72–79.
- ICML-2008-BryanS #learning
- Actively learning level-sets of composite functions (BB, JGS), pp. 80–87.
- ICML-2008-CaronD #parametricity
- Sparse Bayesian nonparametric regression (FC, AD), pp. 88–95.
- ICML-2008-CaruanaKY #empirical #evaluation #learning
- An empirical evaluation of supervised learning in high dimensions (RC, NK, AY), pp. 96–103.
- ICML-2008-CatanzaroSK #classification #performance
- Fast support vector machine training and classification on graphics processors (BCC, NS, KK), pp. 104–111.
- ICML-2008-Cayton #nearest neighbour #performance #retrieval
- Fast nearest neighbor retrieval for bregman divergences (LC), pp. 112–119.
- ICML-2008-CevikalpTP #classification
- Nearest hyperdisk methods for high-dimensional classification (HC, BT, RP), pp. 120–127.
- ICML-2008-ChenM #learning
- Learning to sportscast: a test of grounded language acquisition (DLC, RJM), pp. 128–135.
- ICML-2008-ChenY #kernel
- Training SVM with indefinite kernels (JC, JY), pp. 136–143.
- ICML-2008-CoatesAN #learning #multi
- Learning for control from multiple demonstrations (AC, PA, AYN), pp. 144–151.
- ICML-2008-ColemanSW #clustering #consistency
- Spectral clustering with inconsistent advice (TC, JS, AW), pp. 152–159.
- ICML-2008-CollobertW #architecture #learning #multi #natural language #network
- A unified architecture for natural language processing: deep neural networks with multitask learning (RC, JW), pp. 160–167.
- ICML-2008-Corrada-EmmanuelS #estimation #fault #geometry #low level #precise
- Autonomous geometric precision error estimation in low-level computer vision tasks (ACE, HJS), pp. 168–175.
- ICML-2008-CortesMPR #algorithm
- Stability of transductive regression algorithms (CC, MM, DP, AR), pp. 176–183.
- ICML-2008-CrammerTP #clustering
- A rate-distortion one-class model and its applications to clustering (KC, PPT, FCNP), pp. 184–191.
- ICML-2008-CunninghamSS #estimation #performance #process
- Fast Gaussian process methods for point process intensity estimation (JPC, KVS, MS), pp. 192–199.
- ICML-2008-DaiYXY #clustering #self
- Self-taught clustering (WD, QY, GRX, YY), pp. 200–207.
- ICML-2008-DasguptaH #learning
- Hierarchical sampling for active learning (SD, DH), pp. 208–215.
- ICML-2008-DekelS #learning
- Learning to classify with missing and corrupted features (OD, OS), pp. 216–223.
- ICML-2008-DembczynskiKS
- Maximum likelihood rule ensembles (KD, WK, RS), pp. 224–231.
- ICML-2008-DickHS #infinity #learning #semistructured data
- Learning from incomplete data with infinite imputations (UD, PH, TS), pp. 232–239.
- ICML-2008-DiukCL #learning #object-oriented #performance #representation
- An object-oriented representation for efficient reinforcement learning (CD, AC, MLL), pp. 240–247.
- ICML-2008-DonmezC #learning #optimisation #rank #reduction
- Optimizing estimated loss reduction for active sampling in rank learning (PD, JGC), pp. 248–255.
- ICML-2008-DoshiPR #learning #using
- Reinforcement learning with limited reinforcement: using Bayes risk for active learning in POMDPs (FD, JP, NR), pp. 256–263.
- ICML-2008-DredzeCP #classification #linear
- Confidence-weighted linear classification (MD, KC, FP), pp. 264–271.
- ICML-2008-DuchiSSC #learning #performance
- Efficient projections onto the l1-ball for learning in high dimensions (JCD, SSS, YS, TC), pp. 272–279.
- ICML-2008-DugasG
- Pointwise exact bootstrap distributions of cost curves (CD, DG), pp. 280–287.
- ICML-2008-DundarWLSR #case study #classification #detection
- Polyhedral classifier for target detection: a case study: colorectal cancer (MD, MW, SL, MS, VCR), pp. 288–295.
- ICML-2008-EpshteynVD #learning
- Active reinforcement learning (AE, AV, GD), pp. 296–303.
- ICML-2008-FinleyJ
- Training structural SVMs when exact inference is intractable (TF, TJ), pp. 304–311.
- ICML-2008-FoxSJW #persistent
- An HDP-HMM for systems with state persistence (EBF, EBS, MIJ, ASW), pp. 312–319.
- ICML-2008-FrancS #algorithm
- Optimized cutting plane algorithm for support vector machines (VF, SS), pp. 320–327.
- ICML-2008-FrancLM #fault
- Stopping conditions for exact computation of leave-one-out error in support vector machines (VF, PL, KRM), pp. 328–335.
- ICML-2008-FrankMP #learning
- Reinforcement learning in the presence of rare events (JF, SM, DP), pp. 336–343.
- ICML-2008-GomesWP #bound #memory management #modelling #topic
- Memory bounded inference in topic models (RG, MW, PP), pp. 344–351.
- ICML-2008-GonenA #kernel #learning #locality #multi
- Localized multiple kernel learning (MG, EA), pp. 352–359.
- ICML-2008-GordonGM #game studies #learning
- No-regret learning in convex games (GJG, AG, CM), pp. 360–367.
- ICML-2008-HaffariWWMJ
- Boosting with incomplete information (GH, YW, SW, GM, FJ), pp. 368–375.
- ICML-2008-HamL #analysis #learning
- Grassmann discriminant analysis: a unifying view on subspace-based learning (JH, DDL), pp. 376–383.
- ICML-2008-HeigoldDSN #evaluation #recognition #speech
- Modified MMI/MPE: a direct evaluation of the margin in speech recognition (GH, TD, RS, HN), pp. 384–391.
- ICML-2008-HellerWG #modelling #statistics
- Statistical models for partial membership (KAH, SW, ZG), pp. 392–399.
- ICML-2008-HoiJ #kernel #learning
- Active kernel learning (SCHH, RJ), pp. 400–407.
- ICML-2008-HsiehCLKS #coordination #linear #scalability
- A dual coordinate descent method for large-scale linear SVM (CJH, KWC, CJL, SSK, SS), pp. 408–415.
- ICML-2008-HuynhM #learning #logic #markov #network #parametricity
- Discriminative structure and parameter learning for Markov logic networks (TNH, RJM), pp. 416–423.
- ICML-2008-HyvarinenSH #modelling
- Causal modelling combining instantaneous and lagged effects: an identifiable model based on non-Gaussianity (AH, SS, POH), pp. 424–431.
- ICML-2008-KakadeST #algorithm #multi #online #performance #predict
- Efficient bandit algorithms for online multiclass prediction (SMK, SSS, AT), pp. 440–447.
- ICML-2008-KarlenWEC #scalability
- Large scale manifold transduction (MK, JW, AE, RC), pp. 448–455.
- ICML-2008-KerstingD #parametricity #policy #relational
- Non-parametric policy gradients: a unified treatment of propositional and relational domains (KK, KD), pp. 456–463.
- ICML-2008-KirshnerP #dependence #using
- ICA and ISA using Schweizer-Wolff measure of dependence (SK, BP), pp. 464–471.
- ICML-2008-KlementievRS #modelling #rank
- Unsupervised rank aggregation with distance-based models (AK, DR, KS), pp. 472–479.
- ICML-2008-KohliSRKT #multi #on the
- On partial optimality in multi-label MRFs (PK, AS, CR, VK, PHST), pp. 480–487.
- ICML-2008-KolterCNGD #learning #programming
- Space-indexed dynamic programming: learning to follow trajectories (JZK, AC, AYN, YG, CD), pp. 488–495.
- ICML-2008-KondorB #graph
- The skew spectrum of graphs (RK, KMB), pp. 496–503.
- ICML-2008-KuzelkaZ #estimation #first-order #performance
- Fast estimation of first-order clause coverage through randomization and maximum likelihood (OK, FZ), pp. 504–511.
- ICML-2008-LanLQML #learning #rank
- Query-level stability and generalization in learning to rank (YL, TYL, TQ, ZM, HL), pp. 512–519.
- ICML-2008-Landwehr #modelling #process
- Modeling interleaved hidden processes (NL), pp. 520–527.
- ICML-2008-LangfordSW
- Exploration scavenging (JL, ALS, JW), pp. 528–535.
- ICML-2008-LarochelleB #classification #strict #using
- Classification using discriminative restricted Boltzmann machines (HL, YB), pp. 536–543.
- ICML-2008-LazaricRB #learning
- Transfer of samples in batch reinforcement learning (AL, MR, AB), pp. 544–551.
- ICML-2008-LebanonZ #modelling
- Local likelihood modeling of temporal text streams (GL, YZ), pp. 552–559.
- ICML-2008-Li #approximate #comparison #difference #linear #worst-case
- A worst-case comparison between temporal difference and residual gradient with linear function approximation (LL), pp. 560–567.
- ICML-2008-LiLW #framework #learning #self #what
- Knows what it knows: a framework for self-aware learning (LL, MLL, TJW), pp. 568–575.
- ICML-2008-LiLT #classification #constraints #programming
- Pairwise constraint propagation by semidefinite programming for semi-supervised classification (ZL, JL, XT), pp. 576–583.
- ICML-2008-LiangJ #analysis #generative #pseudo
- An asymptotic analysis of generative, discriminative, and pseudolikelihood estimators (PL, MIJ), pp. 584–591.
- ICML-2008-LiangDK #compilation
- Structure compilation: trading structure for features (PL, HDI, DK), pp. 592–599.
- ICML-2008-LoeffFR #approximate #learning #named
- ManifoldBoost: stagewise function approximation for fully-, semi- and un-supervised learning (NL, DAF, DR), pp. 600–607.
- ICML-2008-LongS #classification #random
- Random classification noise defeats all convex potential boosters (PML, RAS), pp. 608–615.
- ICML-2008-LuPV #analysis #component #multi
- Uncorrelated multilinear principal component analysis through successive variance maximization (HL, KNP, ANV), pp. 616–623.
- ICML-2008-LuLHE #framework #kernel
- A reproducing kernel Hilbert space framework for pairwise time series distances (ZL, TKL, YH, DE), pp. 624–631.
- ICML-2008-MakinoT #network #online
- On-line discovery of temporal-difference networks (TM, TT), pp. 632–639.
- ICML-2008-MartinsFASX #kernel
- Nonextensive entropic kernels (AFTM, MATF, PMQA, NAS, EPX), pp. 640–647.
- ICML-2008-MehtaRTD #automation
- Automatic discovery and transfer of MAXQ hierarchies (NM, SR, PT, TGD), pp. 648–655.
- ICML-2008-MekaJCD #learning #online #rank
- Rank minimization via online learning (RM, PJ, CC, ISD), pp. 656–663.
- ICML-2008-MeloMR #analysis #approximate #learning
- An analysis of reinforcement learning with function approximation (FSM, SPM, MIR), pp. 664–671.
- ICML-2008-MnihSA #empirical
- Empirical Bernstein stopping (VM, CS, JYA), pp. 672–679.
- ICML-2008-KumarT #estimation
- Efficiently solving convex relaxations for MAP estimation (MPK, PHST), pp. 680–687.
- ICML-2008-NarayanamurthyR #markov #on the #process #symmetry
- On the hardness of finding symmetries in Markov decision processes (SMN, BR), pp. 688–695.
- ICML-2008-Nijssen #classification
- Bayes optimal classification for decision trees (SN), pp. 696–703.
- ICML-2008-NowozinB #approach #learning
- A decoupled approach to exemplar-based unsupervised learning (SN, GHB), pp. 704–711.
- ICML-2008-OBrienGG #classification #multi #probability
- Cost-sensitive multi-class classification from probability estimates (DBO, MRG, RMG), pp. 712–719.
- ICML-2008-OrabonaKC #bound #kernel
- The projectron: a bounded kernel-based Perceptron (FO, JK, BC), pp. 720–727.
- ICML-2008-OuyangG #learning #ranking
- Learning dissimilarities by ranking: from SDP to QP (HO, AGG), pp. 728–735.
- ICML-2008-PaiementGBE #distance
- A distance model for rhythms (JFP, YG, SB, DE), pp. 736–743.
- ICML-2008-PalatucciC #classification #on the #scalability
- On the chance accuracies of large collections of classifiers (MP, AC), pp. 744–751.
- ICML-2008-ParrLTPL #analysis #approximate #feature model #learning #linear #modelling
- An analysis of linear models, linear value-function approximation, and feature selection for reinforcement learning (RP, LL, GT, CPW, MLL), pp. 752–759.
- ICML-2008-PuolamakiAK #learning #query
- Learning to learn implicit queries from gaze patterns (KP, AA, SK), pp. 760–767.
- ICML-2008-QiLDC #multi #process
- Multi-task compressive sensing with Dirichlet process priors (YQ, DL, DBD, LC), pp. 768–775.
- ICML-2008-QuadriantoSCL
- Estimating labels from label proportions (NQ, AJS, TSC, QVL), pp. 776–783.
- ICML-2008-RadlinskiKJ #learning #multi #ranking
- Learning diverse rankings with multi-armed bandits (FR, RK, TJ), pp. 784–791.
- ICML-2008-RanzatoS #documentation #learning #network
- Semi-supervised learning of compact document representations with deep networks (MR, MS), pp. 792–799.
- ICML-2008-RavikumarAW #convergence #linear #message passing #source code
- Message-passing for graph-structured linear programs: proximal projections, convergence and rounding schemes (PDR, AA, MJW), pp. 800–807.
- ICML-2008-RaykarKBDR #automation #feature model #induction #learning #multi
- Bayesian multiple instance learning: automatic feature selection and inductive transfer (VCR, BK, JB, MD, RBR), pp. 808–815.
- ICML-2008-ReisingerSM #kernel #learning #online
- Online kernel selection for Bayesian reinforcement learning (JR, PS, RM), pp. 816–823.
- ICML-2008-RenDC #process
- The dynamic hierarchical Dirichlet process (LR, DBD, LC), pp. 824–831.
- ICML-2008-RishGCPG #linear #modelling #reduction
- Closed-form supervised dimensionality reduction with generalized linear models (IR, GG, GAC, FP, GJG), pp. 832–839.
- ICML-2008-Rosset #kernel
- Bi-level path following for cross validated solution of kernel quantile regression (SR), pp. 840–847.
- ICML-2008-RothF #algorithm #linear #modelling #performance
- The Group-Lasso for generalized linear models: uniqueness of solutions and efficient algorithms (VR, BF), pp. 848–855.
- ICML-2008-SahbiARK #kernel #recognition #robust #using
- Robust matching and recognition using context-dependent kernels (HS, JYA, JR, RK), pp. 856–863.
- ICML-2008-SakumaKW #learning #privacy
- Privacy-preserving reinforcement learning (JS, SK, RNW), pp. 864–871.
- ICML-2008-SalakhutdinovM #analysis #network #on the
- On the quantitative analysis of deep belief networks (RS, IM), pp. 872–879.
- ICML-2008-SalakhutdinovM08a #markov #matrix #monte carlo #probability #using
- Bayesian probabilistic matrix factorization using Markov chain Monte Carlo (RS, AM), pp. 880–887.
- ICML-2008-SarawagiG
- Accurate max-margin training for structured output spaces (SS, RG), pp. 888–895.
- ICML-2008-SarkarMP #graph #incremental #performance #proximity #scalability
- Fast incremental proximity search in large graphs (PS, AWM, AP), pp. 896–903.
- ICML-2008-Schnall-LevinCB #algorithm #design #framework
- Inverting the Viterbi algorithm: an abstract framework for structure design (MSL, LC, BB), pp. 904–911.
- ICML-2008-SeegerN #design
- Compressed sensing and Bayesian experimental design (MWS, HN), pp. 912–919.
- ICML-2008-SeldinT #category theory #classification #clustering #multi
- Multi-classification by categorical features via clustering (YS, NT), pp. 920–927.
- ICML-2008-Shalev-ShwartzS #dependence #optimisation #set
- SVM optimization: inverse dependence on training set size (SSS, NS), pp. 928–935.
- ICML-2008-ShiBY #learning #modelling #using
- Data spectroscopy: learning mixture models using eigenspaces of convolution operators (TS, MB, BY), pp. 936–943.
- ICML-2008-ShinK #kernel
- A generalization of Haussler’s convolution kernel: mapping kernel (KS, TK), pp. 944–951.
- ICML-2008-ShringarpureX #named #search-based
- mStruct: a new admixture model for inference of population structure in light of both genetic admixing and allele mutations (SS, EPX), pp. 952–959.
- ICML-2008-SiggB
- Expectation-maximization for sparse and non-negative PCA (CDS, JMB), pp. 960–967.
- ICML-2008-SilverSM #learning
- Sample-based learning and search with permanent and transient memories (DS, RSS, MM), pp. 968–975.
- ICML-2008-SindhwaniR #learning #multi
- An RKHS for multi-view learning and manifold co-regularization (VS, DSR), pp. 976–983.
- ICML-2008-SokolovskaCY #learning #modelling #probability
- The asymptotics of semi-supervised learning in discriminative probabilistic models (NS, OC, FY), pp. 984–991.
- ICML-2008-SongZSGS #estimation #kernel
- Tailoring density estimation via reproducing kernel moment matching (LS, XZ, AJS, AG, BS), pp. 992–999.
- ICML-2008-SorokinaCRF #detection #interactive #statistics
- Detecting statistical interactions with additive groves of trees (DS, RC, MR, DF), pp. 1000–1007.
- ICML-2008-SriperumbudurLL #classification #kernel #metric
- Metric embedding for kernel classification rules (BKS, OAL, GRGL), pp. 1008–1015.
- ICML-2008-SuZLM #learning #network #parametricity
- Discriminative parameter learning for Bayesian networks (JS, HZ, CXL, SM), pp. 1016–1023.
- ICML-2008-SunJY #analysis #canonical #correlation
- A least squares formulation for canonical correlation analysis (LS, SJ, JY), pp. 1024–1031.
- ICML-2008-SyedBS #learning #linear #programming #using
- Apprenticeship learning using linear programming (US, MHB, RES), pp. 1032–1039.
- ICML-2008-SzafranskiGR #kernel #learning
- Composite kernel learning (MS, YG, AR), pp. 1040–1047.
- ICML-2008-SzitaL #approach
- The many faces of optimism: a unifying approach (IS, AL), pp. 1048–1055.
- ICML-2008-TakedaS
- nu-support vector machine as conditional value-at-risk minimization (AT, MS), pp. 1056–1063.
- ICML-2008-Tieleman #approximate #strict #using
- Training restricted Boltzmann machines using approximations to the likelihood gradient (TT), pp. 1064–1071.
- ICML-2008-UenoKMMI #approach #evaluation #policy #statistics
- A semiparametric statistical approach to model-free policy evaluation (TU, MK, TM, SiM, SI), pp. 1072–1079.
- ICML-2008-UrtasunFGPDL #modelling
- Topologically-constrained latent variable models (RU, DJF, AG, JP, TD, NDL), pp. 1080–1087.
- ICML-2008-GaelSTG #infinity #markov
- Beam sampling for the infinite hidden Markov model (JVG, YS, YWT, ZG), pp. 1088–1095.
- ICML-2008-VincentLBM #robust
- Extracting and composing robust features with denoising autoencoders (PV, HL, YB, PAM), pp. 1096–1103.
- ICML-2008-VovkZ #game studies #predict
- Prediction with expert advice for the Brier game (VV, FZ), pp. 1104–1111.
- ICML-2008-WalderKS #multi #process
- Sparse multiscale gaussian process regression (CW, KIK, BS), pp. 1112–1119.
- ICML-2008-WangM #analysis #using
- Manifold alignment using Procrustes analysis (CW, SM), pp. 1120–1127.
- ICML-2008-WangYQZ #analysis #component #composition #feature model
- Dirichlet component analysis: feature extraction for compositional data (HYW, QY, HQ, HZ), pp. 1128–1135.
- ICML-2008-WangYZ #adaptation #kernel #learning #multi
- Adaptive p-posterior mixture-model kernels for multiple instance learning (HYW, QY, HZ), pp. 1136–1143.
- ICML-2008-WangJC #graph
- Graph transduction via alternating minimization (JW, TJ, SFC), pp. 1144–1151.
- ICML-2008-WangZ #learning #multi #on the
- On multi-view active learning and the combination with semi-supervised learning (WW, ZHZ), pp. 1152–1159.
- ICML-2008-WeinbergerS #distance #implementation #learning #metric #performance
- Fast solvers and efficient implementations for distance metric learning (KQW, LKS), pp. 1160–1167.
- ICML-2008-WestonRC #learning
- Deep learning via semi-supervised embedding (JW, FR, RC), pp. 1168–1175.
- ICML-2008-WingateS #exponential #learning #predict #product line
- Efficiently learning linear-linear exponential family predictive representations of state (DW, SPS), pp. 1176–1183.
- ICML-2008-WolfeHK #dataset #distributed #scalability
- Fully distributed EM for very large datasets (JW, AH, DK), pp. 1184–1191.
- ICML-2008-XiaLWZL #algorithm #approach #learning #rank
- Listwise approach to learning to rank: theory and algorithm (FX, TYL, JW, WZ, HL), pp. 1192–1199.
- ICML-2008-YamanWLd #approximate #modelling
- Democratic approximation of lexicographic preference models (FY, TJW, MLL, Md), pp. 1200–1207.
- ICML-2008-YaoL #difference #learning
- Preconditioned temporal difference learning (HY, ZQL), pp. 1208–1215.
- ICML-2008-YuVGS #approach #optimisation
- A quasi-Newton approach to non-smooth convex optimization (JY, SVNV, SG, NNS), pp. 1216–1223.
- ICML-2008-YueJ #predict #set #using
- Predicting diverse subsets using structural SVMs (YY, TJ), pp. 1224–1231.
- ICML-2008-ZhangTK #analysis #approximate #fault #rank
- Improved Nyström low-rank approximation and error analysis (KZ, IWT, JTK), pp. 1232–1239.
- ICML-2008-ZhangDT #algorithm
- Estimating local optimums in EM algorithm over Gaussian mixture model (ZZ, BTD, AKHT), pp. 1240–1247.
- ICML-2008-ZhaoWZ #clustering #multi #performance
- Efficient multiclass maximum margin clustering (BZ, FW, CZ), pp. 1248–1255.
- ICML-2008-ZhuXZ #markov #network
- Laplace maximum margin Markov networks (JZ, EPX, BZ), pp. 1256–1263.

51 ×#learning

20 ×#multi

17 ×#kernel

14 ×#modelling

12 ×#classification

11 ×#analysis

11 ×#performance

10 ×#using

8 ×#linear

8 ×#process

20 ×#multi

17 ×#kernel

14 ×#modelling

12 ×#classification

11 ×#analysis

11 ×#performance

10 ×#using

8 ×#linear

8 ×#process