http://uai2008.cs.helsinki.fi/programme.shtml
UAI 2008
http://uai2008.cs.helsinki.fi/UAI_camera_ready/bo.pdf	36	Greedy Block Coordinate Descent for Large Scale Gaussian Process Regression	academic adaptation adaptive advance advances agony algorithm algorithms analysis annals approach approximate approximation articulated artificial athena augmentation available bartlett based basis beatson belmont bengio bertsekas bhattacharyya billings black bottom bottou brown bulirsch cambridge candela capture chang chapelle cjlin classifier comput computation computer conditional conference csato csie dataset davis decomposition design directions domain down duraiswami efficient elsevier equations evaluation fast feasible formulations forward friedman function gauss gaussian gradient greedy healing herbrich http human humaneva hyperparameters icml ieee improved improvements inference information informative intelligence international interpolation introduction iterative joachims joint journal kanaujia keerthi kernel large lawrence learning least leung library libsvm light line linear machine machines making matching metaxas method methods minimal models motion multivariate murthy neural ninth nonlinear numerical opper optimization ortega pattern platt practical prediction press proceedings process processes processing programming pursuit quinonero radial rasmussen recognition references regression relevance report research rheinboldt rlsc saul scale scholkopf scientific seeger selection sequential several shen shevade siam sigal sindhwani sminchisescu smola snowbird sofware solution sparse speed splines springer squares statistics stoer structured study support synchronized system systems technical through training transform trees trix twin unifying university using variables vector verlag video view vincent vision visual weiss williams without workshop yang york zoutendijk
http://uai2008.cs.helsinki.fi/UAI_camera_ready/otten.pdf	59	Bounding Search Space Size via (Hyper)tree Decompositions	aalborg ability able about accomplished algorithms allows also although american analysis application approximation artificial asymptotic been belief beneficial between bigger blind bound bounding bounds cases clustering clusters combinatorial comparison comparisons complexity computations computer computers conclusions conduct constraint contribution darwiche dbai dechter decomposition decompositions demonstrated dependencies dept depth dermaku desirable deteriorate determinism domains dovgolevskiy dovgolevsky ecai elimination even execution expert exploit faster fine fishelson framework from function functional further future ganzow geiger general genetics give giving gottlob grained graph graphical graphs greatly haplotyping hardness heredity heuristic hope however human hypertree idea ideas improve indeed instances intelligence intelligent into introduce johnson journal kask kaufmann larrosa lauritzen leone likelihood linkage local looking magnitude marinescu mateescu mathematics maximum mcmahan methods models more morgan musliu networks note obtain often online only orders otten overall paper parallel pearl pedigrees personal practical practice presence previously probabilistic probabilities problem problems proceedings proposed prune reasoning references relationship report research respect rough royal rulff samer scarcello scheme schuster science search seems series several showed shown significance silberstein small society some space spaces specification spiegelhalter state statistical stoc structural structure structures subset suitable system systems technical technology that their this thousands tight tighter total tree triangulation tzemach underlying unifying university variable variables very vienna weighted which while width with work zabiyaka
http://uai2008.cs.helsinki.fi/UAI_camera_ready/auvray.pdf	34	Learning Inclusion-Optimal Chordal Graphs	about acknowledgments action algorithms also anonymous approaches approximate approximations attraction auvray bayesian belgian belgium believe binary bioinformatics biomagnet biomod boundary carry case castelo chordal classes combinations comments community concert conditions considering construction context darwiche data decomposition dependency depth designing dimensions direct discussions disjoints distribution distributions driven editors efficient empirical encouraged equivalence equivalent extended faithful figure framework french friedman from fruitful funded further generating genomes graph graphical graphs greedy have hence inclusion inference initiated interest interesting intersection interuniversity investigate investigation isomorph large larger latter layout lead learning lemma markov might model modeling models more neighbourhood network networks numbers obtain office optimal optimality over particular poles policy possible practical presents proceedings programme progress proof properties random recherche references repeating research respect results reviewers same science search september sets side singleton space speaking state steps strong structure structures subsets such suitable supported symmetry thank that their then these this thus transitivity union useful uses variables variational vertex vertices very vincent want wehenkel where which with work would
http://uai2008.cs.helsinki.fi/UAI_camera_ready/antos.pdf	33	Identifying reasoning patterns in games	aaai about action adequately agent agents aids akov albert algorithm algorithms also analyze analyzing arbitrarily artificial bayes behavior behavioral believe blum brian bring brown calculating cambridge capable certain christian christos complex complexity computational computer computing concise conclusion considerable constantinos context continuation cooperation csail daphne daskalakis decision decisions decomposing diagrams different discoverable discovering effectively efficient equilibria equilibrium especially even explaining features first flavor foundation future game games goldberg good graph graphical have heavily help human identify identifying ignorable ignored ijcai important influence information inputs institute intelligence interact interaction jiang journal just kearns kevin koller lead leyton likely littman maid make maker making massachusetts mathematical maximization method michael might milch model modeling models motivations multi mura nash networks normative number obvious other papadimitriou pastime patterns paul pfeffer pierfrancesco polynomial predicting presented promising purposes rational reasoning reasons reciprocity references report representing research rich rigorous ross satinder savings scenarios second shachter shelton singh solving sometimes stoc structured successful such symposium systems talk technical technology that their theory these they third this those thought three time unmotivated used utility vickrey ways when which work
http://uai2008.cs.helsinki.fi/UAI_camera_ready/hui.pdf	49	Toward Experiential Utility Elicitation for Interface Customization	aaai adaptable adapted adaptive adventure affect albrecht allen anal appl astrom bayesian blaylock boutilier brown bunt card chajewska classification claypool conati control data decision driven elicitation emnlp erlbaum estimation formulation foster friendly game getoor goal hillsdale icaps implicit incomplete indicators interaction interest keyhole langlais lapalme markov math mcgrenere mclaren model modeling models moran newell nicholson norman optimal parameter plan play pomdp prediction preference probabilistic problem problems processes psychology recognition references refinement role shahar state statistical support system text translators user utility waseda what with zukerman
http://uai2008.cs.helsinki.fi/UAI_camera_ready/acar.pdf	32	Adaptive Inference on General Graphical Models	above acar acid acids adaptive adding addition alagona algorithm algorithms allowing allows amino analysis annotating applications applied approach approaches approximate artificial atom backbone balanced based basing bayes bayesian belief benchmark between biol biology blelloch both bucket building canutescu case century chain change chem choice choosing computational computations compute computing conclusion conditional conf conformation consistent constant constructed constructs contingency corresponds curr currently cycles data dechter delcher describe directly discrete drawn dunbrack dynamic dynamizing each edges editor effect effects efficient efficiently either elements elimination energy environment estimates estimation example expected experiment experimental experiments expert exploring factor factors faster feasibility february field fienberg figure five folding force framework free frey from function furthermore general generalized gives goal graph graphical graphs grove harper history however hundred ieee ihler implementation incorporated independence inference inform input intelligence intelligent interactions iterative jordan junction just kamisetty kasif kaufman kollman kschischang langmead large larger lauritzen learning libraries likelihood local loeliger logarithmic marginal mateo math matlab maximum mechanical methods mettu model modeling models moderatelysized molecular morgan motion murphy mutations native networks nips node note nucleic opin other over pages particular pearl performed physical possible prediction press previous probabilistic probabilities proc procedure produced product profeta propagation protein proteins provides queries querying rapid real reasoning recomputing references removal represent representation represents research results rigid rotamer rotamers royal running same science scratch scwrl secondary sequential shelenkov show shows siam side significant simply simulation singh size society soda spanning speedup speedups spiegelhalter stat states static statistics steric struct structure structures study sumer support symposium synthetic systems table tables taking test than that their then theory these this time times toolbox trans tree trees treewidth typical typically under unifying updates updating useful using variable variables vittes weiner weiss well where which whose with xing yanover
http://uai2008.cs.helsinki.fi/UAI_camera_ready/duchi.pdf	41	Projected Subgradient Methods for Learning Sparse Gaussians	academic acoustics aspremont association automatic banerjee bertsekas bilmes boyd cambridge conference control convex covariance data factored fitting fortmann gaussian ghaoui goldstein gradient graphical ieee international inverse learning levitin machine matrices method models natsoulis optimization polyak press proceedings processing projection references shalom signal sparse speech techniques tracking trans university vandenberghe
http://uai2008.cs.helsinki.fi/UAI_camera_ready/tarlow.pdf	2	Flexible Priors for Exemplar-based Clustering	adelaide advances affinity algorithms american analysis annals antoniak applications aristotle association august based bayesian between binary blei body boundaries brightness buffet cardinality categories categorization chain classification clique clustering combining computation computational computer conference configurations convex criteria cuts data david degree denoising detect dirichlet diwan dubes dueck efficient efros evaluation exemplar feature february flexible fowlkes frey gatsby gcnu ghahramani givoni golland graphical graphs griffiths gupta hall human iccv icml ieee image indian inference infinite information intelligence international jain jordan journal lashkari latent learning library machine malik markov martin messages methods metric michael mixture mixtures model models mori morris natural neal networks neural neuroscience nips nonparametric normalized objective pages paige passing pattern points potentials prentice priors problems process processes processing propagation radford rand recognition recovering references report sampling sarawagi science segmentation statistical statistics submitted survey systems technical texture transactions unit university unsupervised untangling using variational vision volume welling with workshop wunsch
http://uai2008.cs.helsinki.fi/UAI_camera_ready/singliar.pdf	64	Efficient inference in p ersistent Dynamic Bayesian Networks	about adnan approach asme auai basic bayesian biosurveillance boyen causation chavira compiling complex computational continuous cooper daphne darwiche dash dean denver disease elimination engineering filtering friedman gregory hogan ijcai inference intel john journal kalman kanazawa keiji koller kupferman levander ligence linear local march mark markov michael model networks outbreaks pages persistence prediction press problems proceedings processes reasoning references stochastic stucture thomas time tractable transactions using variable wagner wengkeen william with wong
http://uai2008.cs.helsinki.fi/UAI_camera_ready/hazan.pdf	3	Convergent Message-Passing Algorithms for Inference over General Graphs with Convex Free Energies	above account alently algorithm algorithmic algorithms alternates alternatives american analysis annals approximate approximations argmin argued arguments artificial ascent asserts association athena attains basic bdomain begin belief belmont below bertsekas bethe bilo block blocks bounds bregman bregmans building case cccp censor class closed computation conbex condition conditional conditions conjugate consider constraints constructing continuous convergence convergent converges convex convexity coordinate date dates decompositions define derive described differentiable domain dual duality dykstra dykstras efficient empirical energies energy entropy equal equals equivn estimation every extension factor fenchel find fixed fixing followed following form formulate free freeman frey from funce function functional functions further furthermore generalized geometry global globerson gradient graph graphs heskes ieee ince indicator inference infinity information intel into introduce isfies ithen jaakkola ject jordan journal kappen kikuchi kschischang lagrange lagrangian least ligence linear loeliger loopy lowing mass math mathematical maxb maxima maximize meltzer method methods minb minimization minimize mooij multipliers murphy nction nctions nedi need neumann neural next nonlinear nonstrictly obtains operators opnimal optimal optimality optimization optimize optimizing oriented orthogonal otherwise over ozdaglar pair parai parallel partition points press primal primaldual primalt princeton proceedings product program programming programs propagation proper proving quivalent quote real recall receding recover references regression reich research respect restricted restricting rockafellar rule sati satisfies scheme scientific section seek sequential simplified since solution solve some spaces squares stating statistical strictly studies study sufficient take taken taking that their then theorem theorems theory there therefore this through tion together transactions trees tseng turn uncertainty unconstnained uniqueness university update upper using value valued vector version wainwright weiss whenever which while whose will willsky with yanover yedidia yuille zero
http://uai2008.cs.helsinki.fi/UAI_camera_ready/ficici.pdf	6	Learning and Solving Many-Player Games through a Cluster-Based Representation	aaai action agent agents aligned amer analogy annals anonymous approach approximate approximating approximation arthur assign assigned associated assume asymmetric based bhat brown call cheng cluster clustering commerce compress computation computing conclusions conf daskalakis derived deviation does each econ economic ectation efficient electronic equilibria equilibrium erative eychik focs from full function functions gambit game games graph graphical group groups have hierarchical http importantly incentives individual inductive infinite influence instead interaction into jehiel journal kearns knowledge large learn learning leyton littman lochner machine mansour mathematics mckelvey mclennan mehta method models more nash noncoop number observations opulation original ounded pairs papadimitriou payoff play player players present prevents profiles progress property rationality reasoning reduction reeves references review same singh small smaller software solved sourceforge strategic strategy structure suri symmetric that theory these this through tools tractably treat turocy twin twins twinsymmetric unilateral uses vastly version vorob wellman where which with within
http://uai2008.cs.helsinki.fi/UAI_camera_ready/simma.pdf	63	CT-NOR: Representing and Reasoning Ab out Events in Continuous Time	applications approach arcs artificial assessment automated bahl barham bayesian benhamini berkeley black breese causal chandra classifiers conference constellation continuous controlling cybernetics determining discovery durrett duxbury dynamic edition endence endencies enterprise examples false frequentist friedman geiger geometry goldszmidt graep greenb hayward heckerman herbrich highly hochb host ieee indep inference institute intel international investigation isaacs jaram journal kandula kaufmann koller learned learning ligence ligent lindsay listgarten maccormick machine maltz mathematical microsoft mixture model models morgan mortier multilevel multiple murphy network networked networks nodelman numb oint owerful pearl plausible poissonnetworks practical press probabilistic probability proceedings processes rate reasoning references reliable representation research royal second service services shelton sigcomm simma society spurious statistical statistics structured systems technical testing theory thesis time towards transactions uncertainty workshop zhang
http://uai2008.cs.helsinki.fi/UAI_camera_ready/kim.pdf	22	Feature Selection via Blo ck-Regularized Regression	acknowledgements adjacent affect algorithm american another arbitrary association available based between block burdick causal chain cheung considered context correlation covariates data demonstrated determinants dimensional distance donnelly encodes estimating ewens expression extension fearnhead finding from gene generalized genetic genetics genome george gibbs grants graphical grouped groups high human idea identify information interesting journal learn mapping markers markov mcculloch measurements method model morley mouse multiple nature opulation output paper prior problem proposed rate rates recombination references regional regression relevant represent representing research sampling selection simulated snps space spielman statistical structure subset such supported that there this used variable variables well when wide with
http://uai2008.cs.helsinki.fi/UAI_camera_ready/goodman.pdf	28	Church: a language for generative mo dels	aaai abelson acknowledgements adaptor advances alexey algorithmic annals annual artif artificial authors basis bayesian beal best blaise blog bonawitz brian church clinger combinatorial communication composable compositional computation computer computing concepts conf conference conll constructive csail david definition definitional dirichlet discussions domingos edinburgh editor emnlp expressive extensibility flour formal foundation framework from funded gerry ghahramani goldwater goodman grammars grant griffiths harmeling helpful hidden hierarchical higherorder ibal ijcai independence inductive inference infinite institute intel international interpretation interpreters jects johnson joint jordan kameya kelsey kemp klein knight kolobov laboratory language languages learning liang ligence like logic lunn machine markov marthi massachusetts mathematical mathematics mcallester mccarthy mdps milch model modeling modelling models muggleton natl networks nips nonparametric notes pages part pcfg petrov pfeffer pitman postulates press priors prism probabilistic proc processes programming programs radul raedt random rasmussen rational rees references relational report revised reynolds richardson russell saint sato scheme school sciences semantics sethuraman sinica solving sontag specifying spiegelhalter statistica statistics stochastic storkey structure summer sussman symbolic symbolicstatistical syntactic systems technical technology tenenbaum thank theory thesis this thomas toussaint ueda university unknown using winbugs with work world would yamada
http://uai2008.cs.helsinki.fi/UAI_camera_ready/ganapathi.pdf	20	Constrained Approximate Maximum Entropy Learning of Markov Random Fields	algorithms approach arguments artificial asynchronous available belief berger bethe class cohen computational computer conditional convexity data doctoral domains efficient elidan energies entropy extraction fields finkel free freitag from gibbs gould grenager heskes http icml incorporating informal information informed intelligence international into journal kikuchi koller labeling lafferty language learning linguistics local location logistic machine manning markov maximum maximumlikelihood mccallum mcgraw message minimization minka models multi natural passing pereira pietra prior probabilistic proc processing propagation random references regression relative research residual rodgers sampling scheduling segmentation segmenting sequence stat systems thesis vision with
http://uai2008.cs.helsinki.fi/UAI_camera_ready/jebara.pdf	51	Bayesian Out-Trees	about algorithm analysis annals approach approximating assumptions based bayesian beal being best carreira clustering columbia computer cucs decision density dependent discovery discrimination distributed drineas entropy estimation extremal ferguson forum friedman ghahramani gram graphs helmbold hidden identical improved independent inference infinite information jaakkola jakobson jebara jmlr john kernel koller learning long machine mahoney manifold markov math matrix maximum meila method metrics model nearly network networks neural nips nonparametric nystrom perpinan predicting problems processing proximity pruning rasmussen references report rivin sampling schapire science similarly some song sons statistical statistics structural structure systems technical thadani tiao tree under university well wiley zemel
http://uai2008.cs.helsinki.fi/UAI_camera_ready/hoyer.pdf	48	Causal discovery of linear acyclic mo dels with arbitrary distributions	above acknowledgements acyclic adding additional additionally adjacency affect allowing also although always american among analysis ancestor ancestors anderson andersson annals annual appendix arrows artificial association assume assumed assumptions attention augmented authors auxiliary background basis because been between bollen both build call cambridge cancelled cannot case cases causal causality causation chain change characterization chickering child choice clark classes clear coefficients collect collects collider column columns combination complete completely completes component components concludes conditional conference conflicting connected connecting conservative consider constants containing correct correctly corresponding corresponds could covariance darling decades deduced denot denotes depend dependent descendant descendants differ digraphs direct directly discovery discussed discussions distribution distributions disturbance disturbances dseparation each easily edge edges effect effects either entries entry equal equations equivalence equivalent essentially estimate estimation even exact exactly except explanation expressed faithfulness fashion find first focus follows form from full fully funded furthermore gaussian geiger general generality generally generating given gives glymour goodness grant graph graphs greedy groups happens have heckerman helpful helsinki hence here hoyer identical identically identifiable identification identify ilarly immediate implied implies imply impossible indeed independent independently indeterminancy inference instantiates intel into invertible investigated john joint journal karhunen kerminen know knowing knowledge known lack last latent lead learning lemma ligence like linear looks loss lower machine made madigan making markov match matrices matrix maximal mean means meek method mixed model models more multidimensional must mutually named necessarily need networks newly next ngdag ngdags nodes nondescendants nongaussian notation note noted observed obtain obtained only onto optimal order ordered ordering orient orientation orientations oriented orienting originating other others otherwise over pair paired paper parametrization parent parents pattern patterns pearl perlman permutation possible postdoctoral potentially practical prediction press previously proc proof properly properties property prove provided question ramsey rather reasoning reduced references relative removing represent represented representing represents required research researcher rest restricted result resulting rinen rows same scaling scheines search select selected selecting separately separation separationequivalence setting several share shimizu should show similarly simplicity simply since single solving some somehow sons source span special spirtes springerverlag stand standard statistical statistics stimulating structural structure subsequently such suitable summary terminating terms test than thank that their then theorem there these they this thoroughly thus topic total triangular uncertainty understood union university unshielded used using value variable variables vector vectors weights weigths well what when whenever where wherever which whose wiley will wish with within without would yields zero zeros zhang
http://uai2008.cs.helsinki.fi/UAI_camera_ready/riedel.pdf	61	Improving the Accuracy and Efficiency of MAP Inference for Markov Logic	aaai adaptive advances algorithm algorithms amir anguelov annual artificial bilenko braz carreras chapter clarke compression computational conf conll constraints correlated correspondence crammer dantzig data davis detection discourse discovery domingos duplicate emnlp first fulkerson hundred inductive inference information intel interface intl introduction james johnson joint journal know koby koller labeling language lapata large layer learnable learning ledge lifted ligence logic machine marquez measures mining mirella missing modelling mooney multiclass natural neural next nonrigid online operations order pang pedro press probabilistic problem problems proc processing programming references registration research role roth salesman salvo scale semantic shared similarity singer solution srinivasan string structured surfaces systems task thrun traveling ultraconservative unsupervised using what with years yoram
http://uai2008.cs.helsinki.fi/UAI_camera_ready/grunwald.pdf	45	A Game-Theoretic Analysis of Updating Sets of Probabilities	acknowledgments action addison admissibility algorithmic alternatively always america american analysis annals arguments arxiv asquith association augustin axiomatic axiomatization axioms based bayes bayesian between bliss both calibrated case challenging chapman characterization characterize characterized characterizing community compelling condition conditioning considered contrast convex could cozman dawid decision defined defines depends determine dilation discrepancy distribution distributions divisive economics enquiry entropy epistemic european excellence expected fifty first follows fraassen from full function functions further game gametheoretic gammerman gardenfors generalized gilboa give given good grants graphoid grove grunwald hall halpern have here herron http hughes ignorance imprecise indeed independence instance interesting interpreted irrelevance isipta joseph journal kinematics kitcher know learning least loss magazine making marilyn mathematical maximin maximum maxmin maxprp measure minimax minimaxoptimal minimum mosteller network noting notions observation observed only optimal order ordering paper parade part partial pascal personal perspective peter philosophy philsophical point positive press prior priori probabilities probability problems procedures proceedings programme properties puts random rather reasoning references relative rescher response results right risk robust rule rules sahlin savant schmeidler science scientific second seidenfeld select sense sept sets shafer shows similar since single solutions somehow special springer statistical statistics strategy suboptimality supported symmetries symmetry synthese take taking than that their then theorem theoretic theory thing this those uncertainty under understood unique university unreliable update updating used utility versus view viewed volume vovk wald walley wasserman well wesley when whether wiley with world worth would
http://uai2008.cs.helsinki.fi/UAI_camera_ready/nielsen.pdf	15	Explanation Trees for Causal Bayesian Networks	aaai abduc abnormal absent algorithm algorithms allowing allows application approach approximation asia assuming available based bayes bayesian belief biometrika british bronchitis build cambridge campos cancer castilla causal causality causally causation causes chan changes compactly compare complexity computations conclusion condition copenhagen criterion cusing darwiche decision dening diagnosis diagrams diez distinction druzdzel dyspnea easy ecause edition empirical ence engineering ensure etween expert explain explaining explanandum explanation explanations explanatory explicit exploiting factor favorably figure finding flores formation frey fuzziness given glymour graphical graphs halpern have henrion ieee increase independence infer inference information input instead institute intel international interpret inthis intuitive inuence irrelevance jensen jereys jewska journal kauf kaufmann know knowledge kschischang lacave lauritzen leads leaves ledge lets ligent list local loeliger long lung makes making mancha mann master mathematics maximize means methods model models moral more morgan most network networks nielsen noncausal normal observation observed only osed ositive ossibly other oxford pages park part partial pearl philosophy planck plausible polani prediction present presented press principles probabilistic probabilities probability probable product prop propagation qualitative readable reasoning reect references report representing research result results returned review robustness royal scenario scheines scheme science sciences search second selected sensible series several shimony simplifying sity smoker society spiegelhalter spirtes springer state statistical still structural structures systems tborca technical techniques tems tested than that their them theory thesis they this those tion transactions tree trees triangulation tuberculosis uncertainty univer universidad university unlike user vanced variables visit while with workshop yuan
http://uai2008.cs.helsinki.fi/UAI_camera_ready/yu.pdf	71	Refractor Imp ortance Sampling	aaai absorption academy according afit alarm alberta algorithms also analysis anytime appendix application approach approaches approximate approximating architecture artificial bandwagon bayesian beinlich belief between biocybernetics biomedical bypasses canada cand carlo case cases causal chain chang chapman chavez choice company computer conference configuration consider constant constructing cooper corollary cross decided decision deliberative department descendants diagnosis difference disorders distributed distribution divergence does each editor editors edmonton electrical elsevier engelen engineering entropy equal esbn evidence evidential exist exists expert first following follows force fourth from fung further gaag garvey geman general gibbs gilks graphical hall have hence henrion hoboken holland ieee images infc inference information institute integrating intel intelligence international introduction issues john joint jordan journal kanal kaufmann kullback learning lemma lemmer lesser levitt ligence ligent liver logic london machine mackay markov mateo measure medicine method methods minimize minimizing models modern monitoring monte morgan must neapolitan network networks north norvig onisko only order original pages parents pass path paths pattern pearl peot plausible polish possibilities possible posterior practical practice prentice press probabilistic probability proceedings propagating proves publishers publishing real reasoning references relaxation removal report research restoration richardson rmation rubinstein russell sampling santos science second separate separated separates series shachter shannon shimony should show simplified simulation since solomon some sons spiegelhalter statistics stochastic study suermondt support survey system systems technical techniques technology term terms that then theorem theory there thesis third this three thus time topological transactions uncertainty using volume warsaw weighting where wiley williams with workshop york zero
http://uai2008.cs.helsinki.fi/UAI_camera_ready/cseke.pdf	38	Bounds on the Bethe Free Energy for Gaussian Networks	advanced advances aistats analysis approximate approximation artificial bayesian becker belief bethe cambridge conditional conference constraints cowell december divergence editors empirical energy expectation experiment field fields fifteenth fixed francisco free gaussian ghahramani graphical heskes horn http inference information intel jaakkola johnson jordan journal jstat kaufman kaufmann learning ligence ligent loopy machine malioutov mateo matrix mean measures mechanics message methods microsoft minima minka models morgan murphy networks neural obermayer october opper pages passing pearl plausible points power practice press probabilistic proceedings processing propagation publishers random reasoning references report research robert saad stable statistical study sums systems szummer technical techniques theory thrun tutorial uncertainty university variational walk weiss wiegerinck willsky winther with zoeter zoubin
http://uai2008.cs.helsinki.fi/UAI_camera_ready/mimno.pdf	0	Topic Mo dels Conditioned on Arbitrary Features with Dirichlet-multinomial Regression	about accounted acknowledgments additional adds agency allocation also annotated area assignment attributes author authors beta better between beyond bickel blei calculation center central chemudugunta citation complement complicated conditioned connection constructed continuous contrast converges corrada correlations dagstructured data date decreases density dietz dirichlet dirichletmultinomial discovery discussions document documents econometrics econwpa emmanuel empirical entity equivalent erosheva especially expertise faster features fienberg figure found from future generated generative grant griffiths group guimaraes hall hanna have helpful hidden hierarchical http hybrid icml ijcai incorporated inferences influences information intelligence intelligent into iterations january jordan journal lafferty language large latent lavine learning likelihood limited lindrooth machine made make mallet matching mathematical mcauliffe mccallum membership memory metadata method michael mimno mixed mixture model modeling models mohanty more must nanotech national networks newman nips nocedal nonmarkov number observed optimization optimizations outperforms over overhead pachinko papers parameter parameters part perplexity phase pnas pointing prediction programming provides publications references regression relations research result retrieval reviewers role rosen roughly sampled sampler sampling scale scheffer science scientific security sept shows sigir simple slda smyth social specific splitting statistical steyvers substantially such supervised suppl supported term than thanks that their this time toolkit topic topical topics trained training trends umass unsupervised useful variables wallach wang which while with words work xuerui
http://uai2008.cs.helsinki.fi/UAI_camera_ready/kveton.pdf	13	Partitioned Linear Programming Approximations for MDPs	allocation approach approximate approximation artificial athena bella bellman belmont benjamin bertsekas bertsimas boutilier computation computational conference construction craig daniela dearden dimitri dimitris dynamic exploiting farias goldszmidt intelligence international introduction john joint kalaba kotkin linear mathematics moises neuro operations optimization pages policy polynomial press princeton proceedings processes programming pucci references research richard robert scientific structure technique tsitsiklis university
http://uai2008.cs.helsinki.fi/UAI_camera_ready/radovilsky.pdf	60	Observation Subset Selection as Lo cal Compilation of Performance Profiles	boddy dean dependent ijcai pages planning problems references solving time
http://uai2008.cs.helsinki.fi/UAI_camera_ready/vanhatalo.pdf	67	Modelling lo cal and global phenomena with sparse Gaussian pro cesses	accurate acknowledgments additive advances agency algorithm also amestoy analysis application approximate approximating approximation approximations artificial author authors automation bayesian because becker behavior better bishop both cambridge candela certain chen circuit class college combines coming comments compact compactly compared computational computer computing conclusions conditions conducted conference consists correlation covariance data davis degree deisenroth delve desc dietterich direct discontinuities duff editors efficient electronics engineering environmetrics equally excellent fagan fast faster field finnish first flexible formation forward found frey full function functions funding gaussian geta ghahramani gives global gneiting good graduate hierarchical hopfield html http ieee impedance industry infeasible inference information informative innovation inputs intel intelligence international introduced inverse journal kernel large lawrence leads learning leen length less ligence linear local london long machine machines marc markov martino mathematical matrix memory method methods minimum model modeling modelling models more multivariate nature networks neural niessner ninth nonadditive nonstationary numerical nystr obermayer often ordering overall paciorek paper performance phenomena phenomenon planning platt poor post power practical press problems proceedings process processes processing properties proposed provides pseudo purposes quinonero random rasmussen rather reasonable references regression reichert requires research routines scale scales scattered schervish schlkopf school seeger selection sets several short shown siam similarly snelson society software sparse spatial speed statistical statistics storkey structure studies study support supported systems takahashi technology telecommunications than thank thanks that there thesis this thrun time tool toronto training transactions tresp under unifying university used using vector view volume weiss well wendland which williams with workshop
http://uai2008.cs.helsinki.fi/UAI_camera_ready/ganchev.pdf	9	Multi-View Learning over Structured and Non-Identical Outputs	adaptation adaptive aggregating agree algebraic altun analysis ando approach assessments balcan bartlett based behavior belkin berlin bhattacharyya biographies blenders blitzer blum bollywood boom boosting bordley boxes brefeld buchholz buscher canonical cardie chapelle chunking classes classification closed cohn collins colt combining communications complexity computation conditional conll constraints coregularization correlation correspondence data datasets decompose decomposes desired discriminative distance divergence domain dredze dual each emnlp entity entry equation expectation features fields filter form formula foster framework from fujino functions furthermore ganchev generative ghani gives graca greiner have heidelberg hidden hofmann hybrid icml ieee improved independent into introduction investigating isozaki jiao johnson journal kailath kakade kernel know label labeled labeling lang language large learning letting limitations lkopf logarithmic loss machine management mann markov maximization mccallum mcdonald measures meila methods meulder mining mitchell model models morristown multi multiple multiplicative naacl named natural netnews newsweeder nigam nips niyogi noting opinion optimization osborne osition output perceptrons pereira performing pierce pools possible posterior predictive press probability problem proc proceedings program prop proposition rademacher random recognition references regression regularization regularized require research result robust rosenberg sang scalable scheffer schuurmans science segmentation selection semi semio sentiment sequence sequences shared shen sigdatemnlp signal simple simplification sindhwani singer smith solution springer stats structural structured structures style substituting supervised suzuki taking task taskar tasks text that theory this training transactions understanding unlabeled unsupervised vector view views wang where which with workshop yields zhang zien
http://uai2008.cs.helsinki.fi/UAI_camera_ready/duong.pdf	42	Knowledge Combination in Graphical Multiagent Models	about accurate address advances agent agents aggregation algorithm algorithmic alone also alternative although american among analysis another applications approach approaches approximation arbor articial association availability based basic behavior belief beliefs boltzmann boost bounded bridge cambridge canonical capturing characterizations combination combining come commerce complex computational computing concepts conclusions conference connecting daskalakis data dataand decision demonstrated design develop diagrams dierent direction directly distributions does eciency economic editors eective eectively either elds electronic employing engineered eort equilibria especially essential estimation exibility exible existence expect exploiting explored exten extended facilitates feasibility formalism found framework freeman from functions furthermore game games generalized gmms gneiting graph graphical graphically groups guez have help heuristic history html http identied important improve industry information input intel intelligence internet inuence issue itself jaakkola jordan journal kappen kearns knowledge koller krebs large learning letters ligence littman logical machines making markov match mean mechanism method milch mixing mkearns model modeling models most multi multiagent multiple nash nding netindustry networks neural nevertheless nips nisan number observable only orgnet other outcome over pages papadimitriou paper parameters partially partnerships pattern pennock performance pfeer physics play position possibility potential power practical predicting prediction predictive presented press probability problems process processes processing propagation proper provide pure qualitatively raftery random rational reasoning recognition references regret reinforcement representation representing research responsive risk rodr roughgarden rules scalability scenario scenarios scoring seattle seventeenth seventh sharing singh situations solution solving source sources specication springer statistical step strategic strictly structure structured study such support supporting systems tardos task techniques than that theoretic theory this those threshold topology triedmixing tutorial uncertainty university upenn vazirani vorobeychik weiss well wellman what with wolpert workshop yedidia
http://uai2008.cs.helsinki.fi/UAI_camera_ready/ross.pdf	62	Model-Based Bayesian Reinforcement Learning in Large Structured Domains	acting adaptive algorithms amherst analytic andre approach approximate artif artificial barto based bayes bayesian being belief bellman boutilier cassandra chickering combination computational data dearden decision dense discovery discrete distribution domains duff dynamic eaton efficient empirical error exploration factored figure friedman full geiger goldszmidt guestrin heckerman hoey icml intel introduction joint journal kaelbling knowledge known koller learning ligence linear littman machine markov markovian massassachusetts mathematics mcallester mcmc mdps mechanics model murphy network networks number observable optimal parr partially planning pomdps poupart press procedures process processes programming references regan reinforcement representations research return simplification singh solution state statistical steps stochastic structure sutton ternary thesis tree university using venkataraman vlassis with
http://uai2008.cs.helsinki.fi/UAI_camera_ready/decooman.pdf	40	Sensitivity analysis for finite Markov chains in discrete time	artificial artint based bridging campos chains college computing cooman costa decision dimuro editors enrique essays filip finite fuzziness generalizing gert harmanec harper henry hermans honor huete imprecise inference intelligence international intervals intervalvalued itoh january john journal kemeny king knowledge kreinovich kyburg laurie markov miranda models moral nakamura observable pages parameters partially paso planning predictions press probabilities probability processes publications reasoning references report rocha snell stationary statistical step symmetry systems technical texas theories tool trees uncertain uncertainty university utep versus wheeler with
http://uai2008.cs.helsinki.fi/UAI_camera_ready/lacerda.pdf	19	Discovering Cyclic Causal Mo dels by Indep endent Comp onents Analysis	algorithms analysis applied best bollen chegireddy comon component concept discrete equations finding hamacher independent john latent matchings mathematics perfect processing references signal sons structural variables wiley with york
http://uai2008.cs.helsinki.fi/UAI_camera_ready/wexler.pdf	70	Inference for Multiplicative Mo dels	addition advantage algorithm analysis annals artificial automatically base bayesian bipartite bishop bles bottom boyd burke called cambridge captures causal cause classifier comprised computational computing conditional contain contains context contextual contingency contingent cybernetics decision dence dependence dept developed diagnosis diagnostic diagrams dimensional discrete diseases disorder does elements engineeringeconomic entropy especially evaluation every example exploiting extension factor fienberg finding findings fisher formulation found fourth frey function functions fung gains geiger generalized good graph graphical graphs heckerman holland holtzman hypothesis identical ieee indein indepen independence inference influence inform inshop intel international internist jair journal kauffor knoke knowledge kschischang landgrebe lauritzen level ligence ligent lindsey linear loeliger macmillan mann math matheson mations maximum mentioned methodology methods middleton model models morgan multi multidimensional multinets multiple multiplicative nary negative netdiagnosing network networks nodes noisy noisypress oliver oper order oxford pages paper parents part pearl pendence performance philosophical plausible poole potts press probabilistic probabilities proceedings product proposed publications quality quantity quickscore reasoning references reformulation relationships representation represented research running safavian sage same savings scientific section shachter siam similarity smith society some speed stanford stat statistical statistics structrees structures structuring substantial such survey syss systems taan tables takes tems that theory there tivariate tractable trans transactions transfori tree university uses using variables when where which with work workers working workture zhang
http://uai2008.cs.helsinki.fi/UAI_camera_ready/niepert.pdf	29	On the Conditional Indep endence Implication Problem: A Lattice-Theoretic Approach	about addition algo algorithm algorithmic algorithms already annals approach approx approximations artificial ascending based bouckaert bounds characterization combinatorial combined complete complexity conclusion conditional conference constraints could criterion database dence derived descending differential efficiently evidence exclusion experimental falsification future gaag geiger graphical gucht have heuristics implication implications imsets indepen independence independencies infer inference inferences information insights instance instances intel into kaufmann kullback larger lattice leibler ligence ligent like logical mathematical matu models more morgan much networks number pages paper part pearl plausible polynomial possibly prague preprocessing presented principles probabilistic problem proceedings properties provide publishers racing reasoning references related relations reported representation rithms sayrafi scales search shrink space springer stable statements statistics step store structural structures studeny sufficiency symposium system systems than that theory this those time transactions uncertainty used utilized valid verlag waal which with work
http://uai2008.cs.helsinki.fi/UAI_camera_ready/huang.pdf	31	Cumulative distribution networks and the derivative-sum-pro duct algorithm	algorithm american analysis applications arco artificial bayesian buntine carlin chain chapman chess conference data edition factor fields frey gelman ghahramani graepel graphical graphs herbrich ieee inform information intel introduction jaakkola jordan kinderman kschischang learning ligence loeliger machine markov mathematical methods minka models neural nips past players present proceedings processing product publishing random rating references rubin saul science second skill snell society statistical stern system systems texts their theory trans trueskillt uncertainty variational york
http://uai2008.cs.helsinki.fi/UAI_camera_ready/kuroki.pdf	52	The Evaluation of Causal Effects in Studies with an Unobserved Exp osure/Outcome Variable: Bounds and Identification	acknowlegdement addition american analysis applications approach association balke based bayesian between biometrics biometrika birth bollen bounds bross categorical causal chilton clinical cognitive compliance computation considerations contingency counterfactuals criteria culture dental derived difficult education effects empirical enable equations evaluate exist exposure finch fleiss foundation from geiger goodman gradients graphical health identifiability identifying imperfect independence japan john journal kurata laboratory latent mazda method ministry modeling modified multi multidimensional networks observational observe others outcome paper partly path pearl policy posterior probabilistic program provided proxy references report research results ridit science semantics services socioeconomic some sons sports statistical structural studies supported systems tables technical technology this tightest treatment ucla unmeasured unobserved using variable variables verma wallenstein weight when wiley with
http://uai2008.cs.helsinki.fi/UAI_camera_ready/cussens.pdf	39	Bayesian network learning by compiling to weighted MAX-SAT	aaai about accounting adrian alarm alternative american approach association assuming bart bayesian bdeu beame being bram carpo castelo chal chickering cliques cohen coloring column combination conclusions consider counting daphne data dataset david dependencies deterministic differ different dimacs discovery domingos driven dynamic editors efficient encoded encodings estimates even exploration exponential family features figure first friedman from future geiger given graphical greater heckerman henry here hoifung hyderabad identifier ijcai implementation inclusionas incorporation inference integer johnson journal kautz knowledge koka koller learning lenge local machine madigan maxwalksat method michael million model models more move needs negated network networks number occam only orderings pages parent particular paul pedro pittsburgh poon posterior preliminary prior priors probabilistic probability probable problems proc raftery references regular research results robert rounded runs sang satisfiability score search second seen selection selman sets should solving sound space statistical strategies structure structures such testing than they though tian times trick uncertainty uniform used using variable weighted whose window with work worth
http://uai2008.cs.helsinki.fi/UAI_camera_ready/kwisthout.pdf	53	The Computational Complexity of Sensitivity Analysis and Parameter Tuning	aalborg analysis annals approximate architecture artificial auai bayesian belief beliefnetwork bounding campos castillo chan change charitos complexity computational conference coup cozman credal cybernetics darwiche dbns decision discrete distance edinburgh engineering fabio from gaag guti habbema hadi ieee inferential intel international jensen joint journal know ledge ligence making mathematics measure multiple networks pages parameters press probabilistic properties quantification reasoning references report review rrez rulff sene sensitivity single sitivity systems technical threshold transactions twentieth uncertainty university with
http://uai2008.cs.helsinki.fi/UAI_camera_ready/noto.pdf	21	Learning Hidden Markov Mo dels for Regression using Path Aggregation	aaai academy acids advances algorithms alignment altschul analysis annual applications approaches architecture artificial aspects bailey beer bengio biocomputing bioinformatics biol biology biopolymers boguski botstein bratko brown browser burge cambridge carmel cells changes cismodule classification complete comput computational conditional conf conlon craven data detect detecting diekhans discover discovery discriminative editors eighteenth eighth eisen elements elkan environmental eskin expectation expression fields first fisher foundations frasconi friedman from furey gasch gene genome genomic genomics gibbs harel hastie haussler hidden hierarchical hinrichs homologies human identifying ieee information input integrating intel intelligence international jaakkola journal karali karlin karolchik kaufmann kent kernel kramer krogh labeling lafferty language lawrence learning leen leslie lieb ligent logical machine manning markov maximization mccallum method mian mixture model modeling models modules molecular morgan motif motifs multiple national natural neural neuwald noble noto novo nucleic order output pacific pages pereira plos predicting prediction press probabilistic proc processing programs protein rabiner random recognition recomb reexamination references regression regulatory remote represent research response retrieval roskin sampling schutze science sciences segal segmenting selected sequence seventh sharan shen signals sjolander spatial spectrum speech spellman springer statistical storz strategy string structural structures subtle sugnet symposium systems table tavazoie tesauro that thirteenth tibshirani tistical tompa tool touretzky trees tutorial ucsc unsupervised using verlag volume wide wong wootton workshop yakhini yeast york yuan zhou zilberstein
http://uai2008.cs.helsinki.fi/UAI_camera_ready/miller.pdf	58	The Phylogenetic Indian Buffet Pro cess: A Non-Exchangeable Nonparametric Prior for Latent Features	about achieve advances aistats also artificial aspects bayesian beal before behavior below berkeley beta between blei block buffet carlo casella chancellor checked choice close clustering coalescents columns comparison computers conclusion conference correct corresponding data daum defining department described details diaconis different diffusion dirichlet distributions does dropped economic elimination estimate estimated examples exchangeability exchangeable experimental explain faculty feature features figure finally finetti foundation from function fund ghahramani glomerative grant greeno griffiths have hierarchical icml immediately indian individual inference infinite infinitely information instruments intel international jects jordan journal kaufmann knowledge laboratory large latent lawrence learning lectures ligence ligent livermore luce machine many mathematical matlab matrices matrix mcfadden mean methods model models monte more morgan national neal necessary need needed networks neural nips nobel notions novel number numbers observations observe observed pair parameters partnership pearl performance phylogenetic pibp plausible posterior preference presented prior priors probabilistic probabilities probability proceedings process processes processing progress psychological psychology rasmussen reasoning recent recovers references relationships report requires research restle review robert rumelhart same samples sattath schmid science sciences seen show similar similarity springer statistical statistics still stimuli structure summarize supported systems technical test than theory thibaux this though toronto tree trees true tversky university using values very weight weights when while whose wickelmaier wiley with within work
http://uai2008.cs.helsinki.fi/UAI_camera_ready/steck.pdf	18	Learning the Bayesian Network Structure: Dirichlet Prior versus Data	akaike ambrosio bayesian buntine caski extension information international likelihood maximum networks pages petrox principle references refinement second symposium theory
http://uai2008.cs.helsinki.fi/UAI_camera_ready/streeter.pdf	65	New Techniques for Algorithm Portfolio Design	aimath algorithm algorithmica aloul apao arathi avrim blum cover external fadi feige from gagliolo generic iccad igor internal journal jurgen karem learning machine mansour markov matteo namic pages portfolios prasad proximating ramani references regret research sakallah schmidhuber specialized tetali update uriel versus yishay
http://uai2008.cs.helsinki.fi/UAI_camera_ready/hines.pdf	47	Learning When to Take Advice: A Statistical Test for Achieving A Correlated Equilibrium	adaptive advice after although always american analysis artificial association aumann balakrishnan behave behavior behavioral boston calibrated case cohen colell combining commerce compared conference consider constant continuous correlated correlation denoted distributions during each econometrica economic economics edition electronic ended environments equally equation equations equilibria equilibrium equivalent expectation expected expert farias figure finite first fixed flexible following foster free frequency functions galeotti games goyal graphical graphically greenwald hall hart hold holds however hypothesis icml index intelligence jackson johnson journal kakade kearns kotz langford leading learning lemma limit limj linear mathematical mediated megiddo monderer more must nash necessary network note order ortiz pages period play power procedure proceedings proof prove proving randomized reactive references relationship repeated representation represented respect same sciences shown simple simplicity since specifically springer statistical statistics strategies strong subjectivity such sufficient tennenholtz testing than that therefore this time transformations underrepresented univariate unpublished utility vegaredondo vohra volume washington wasserman where will with yariv york young
http://uai2008.cs.helsinki.fi/UAI_camera_ready/hansen.pdf	46	Sparse Stochastic Finite-State Controllers for POMDPs	aaai amato approximate artificial athena bartlett based baxter bernstein bertsekas bounded carin cassandra conference constrained decentralized estimation european examples factored feng file hansen horizon http ijcai index infinite intelligence international introduction iteration joint journal liao linear national optimization page parr planning point policy policygradient pomdp pomdps proceedings programs quadratically references repository research scientific searching shtml solving space tony tsitsiklis uncertainty using zilberstein
http://uai2008.cs.helsinki.fi/UAI_camera_ready/lawrence.pdf	55	Improving Gradient Estimation by Incorp orating Sensor Data	artificial bartlett baxter estimation horizon infinite intel journal ligence policygradient references research
http://uai2008.cs.helsinki.fi/UAI_camera_ready/christoudias.pdf	10	Multi-View Learning in the Presence of View Disagreement	accuracy acoustic across activity adaptation adaptive agreement aistats analysis analyzing ando applicability applications approach association audio automatic averaged bars based bayesian belkin bickel bipartite blum canonical chapman christoudias classiers classification classifiers clustering clustring collins colt combining concept conference contrast conventional coregularization corpora correlation cotraining cover cross cvpr darrell dasgupta data dataset density detection deviation dhillon dimensional disagreement discovery displayed documents dynamic each earance ectral edition edtion eech effectiveness eled elements empirical endency endent entity error erstadt ervised estimating estimation european feature figure first fisher foster generalization generation gesture ghani graph halb hall heterogeneous high human icassp icmi icml ihler indep inference information international into jects joint june kakade knoblock know krishnapuram labeled language large learning ledge levels littman local machine mallela management mcallester measurements methods mining minton mitchell mixture modal model models modha morency multi multiple muslea named naphade natural nigam ninth nips niyogi novemb oracle orly otstrapping ounds over pages partitioning percent performance performs presence probabilistic proceedings processing rahimi random recognition references regression rosales saenko scheffer second seed semantic semi show sigdat sigkdd significance significant silverman sindhwani singer siracusa sons speech splits statistics steck step supervised test tests theoretic theory thesis thomas tieu towards train training unlab unlabeled unsup urtasun user using validation very video videos view views visual well wiley willsky with words workshop york zhang
http://uai2008.cs.helsinki.fi/UAI_camera_ready/meila.pdf	57	Estimation and Clustering with Infinite Rankings	addison advances algorithm american analysis analyzing appear approach artificial association based berlin bilmes buhmann busse cambridge carreira cluster clustering cohen college combinatorial conditional conference consensus critchlow data degroot department distance dublin editors estimation exploring exponential fligner gaag gaussian gormley heidelberg heterogeneity heterogeneous icml ieee infinite information integer intel international irish journal lafferty learning lebanon lecture ligence machine mathematische mean meila methods metric mixture model modelling models multistage murphy nemhauser neural notes number optimization orbanz order page parr partial pattern patterson perpinan phadnis poset press probability proceedings processing rank ranked ranking rankings reading references report research royal schapire shift singer society springer statistical statistics systems technical things thoma tokyo trans trinity uncertainty under verducci verlag volume voting wesley wiley with wolsey york zeitschrift
http://uai2008.cs.helsinki.fi/UAI_camera_ready/gruber.pdf	1	Latent Topic Mo dels for Hyp ertext	academy according achieves additional advances advantage aggregation aistats aleks allocation american amit among analysis andrew annual applying approaches artificial aspect author authoritative authors average bagof based because been between beyond bickel blei blogs bottou brin bringing buntine cambridge cases chang chickering citation cohen cohn compared comparison compui computation compute computed conf conference connectivity content corpus correlated created data david deerwester definition denote depends details dietz different digital dirichlet discrete discussion distribution document documents dumais dynamic each easily editors efficient efficiently endix environment equation erosheva except existence expectationpropagation expectations expected fail fienberg figure first fitting follow formation francisco frequency from full furnas generalization generalize generation generative george given good griffith griffiths gruber halpern hanna harshman have hidden hofmann huan hyperlinked hyperlinks hypertext icml identify importance indexing influence influences infore information innerdbrackets integrating intel intelligence international introduces jakulin ject jeev john jordan joseph joshua journal june kaufmann kleinberg lafferty landauer latent latter laura lawrence learning least library ligence link links lthm machine mark markov mation measure media method methods michael michal minka missing mixedmembership model modeling models morgam morgan motwani much nallapati national neural novel number observation observations order other outperform outperforms overfit overfitting padhraic page pagerank pages parameter parameters pennsylvania percentage pittsburgh plsa plsalda possibly posterior precision predicted prediction presented press previous previously probabilistic probabilistically probabilities probability proc proceedings processing publications publishers ramesh ranking recall references relative report requires research result results richard rosen saul scheffer science sciences scientific scott semantic sergey shown size smaller smyth social society source sources stanford statistics steffen steps steyvers suggests susan symmetries syntax systems taking target tation technical technologies tenenbaum terry test text than that then there these this thomas three times tobias topic topics train true uncertainty under unsupervised wallach webkb weblogs weiss well where which while wikipedia william winograd word words work worse wray written yair
http://uai2008.cs.helsinki.fi/UAI_camera_ready/bhattacharjya.pdf	35	Sensitivity analysis in decision circuits	acknowledgements addison adnan advances analysing analysis anonymous answers applications approach arlington artificial assessments auai banff bayesian behavior belief bhattacharjya boutilier breese building california cambridge canada castillo chan chavira chichester chickering choices circuits compiling computationally conference cybernetics darwiche data december decision decisions diagrams differential discrete dynamic economic edinburgh edition editors edwards efficient energy evaluating evolution explanations factoring feedback ference forum foundations francisco from funded gaag games goldszmidt group gutierrez hadi halpern helpful howard humans ieee inference influence information intel international introductory jensen john joint journal kaufmann kjaerulff know koller laskey lectures ledge ligence local logical making march mark mateo matheson matter menlo miles model modeling morgan morgenstern most multiple nets networks neumann nielsen nineteenth november numbers operations paper parameters park parr part partially press princeton principles probabilistic probability probable proceedings programming providing queries raiffa really reasoning references relevant renooij representation research reviewers robustness savage science scotland seattle second sensitivity shachter single sixteenth stanford statistics strategic structure support systems thank their theory third this transactions twentieth twenty uncertainty under university value vancouver virginia wesley weyant when wiley winterfeldt with working york
http://uai2008.cs.helsinki.fi/UAI_camera_ready/toussaint_revised.pdf	14	Hierarchical POMDP Controller Optimization by Likeliho o d Maximization	aaai algorithms amato approximate automated bernstein boutilier braziunas brown cassandra charlin computer constrained controllers decision dept discovery exact hierarchy ijcai linear local markov observable pages partial planning pomdp pomdps poupart processes programs quadratically references science search shioda solving stochastic thesis university using zilberstein
http://uai2008.cs.helsinki.fi/UAI_camera_ready/el-hay.pdf	27	Gibbs Sampling in Factorized Continuous-Time Markov Processes	approximate bayesian berlin chains chung conference continuous friedman inference koller kupferman markov networks probabilities proceedings references sampling second shelton springer stationary time transition twenty uncertainty verlag with
http://uai2008.cs.helsinki.fi/UAI_camera_ready/isaza.pdf	50	Speeding Up Planning in Markov Decision Processes via Automatically Constructed Abstractions	abstraction abstractions acknowledge acknowledgements action actions advantages aiide alberta algorithm approach approximately artificial asadi athena automatically baldur belmont bertsekas between bioware bjornsson blizzard boutilier building bulitko chaos comments computing conclusions constructed control corp council dean decision demonstrate difficulty dynamic effective effectively efficient empirically engineering entertainment exact existing explored factored fast finally flairs framework fund funded gate givan goal gordon graph gratefully greiner hauskrecht heuristic hierarchical hierarchies homogeneous http huber icaps icore ijcai independent ingenuity insightful intelligence isaza jair july kaelbling knowledge leach learning lustrek macro markov mcmahan mdps memory meuleau model national neuro november nserc optimal paper part partitions path pathfinding planning precup press prmdp problems processes programming provide real reduction references reign reinforcement representation research results reviewers schaeffer science scientific search semi showing sigmundarson singh size skill solution solutions solving space speed speeding ssps state strengthen sturtevant sutton szepesva szepesvari tech techniques temporal that then theoretical this through time transfer tsitsiklis ualberta using varying warcraft ways with works
http://uai2008.cs.helsinki.fi/UAI_camera_ready/munoz.pdf	7	A Polynomial-time Nash Equilibrium Algorithm for Rep eated Sto chastic Games	arti bellman bowling conference convergent dynamic games international learning press princeton proceedings programming rational references seventeenth stochastic university veloso
http://uai2008.cs.helsinki.fi/UAI_camera_ready/lowd.pdf	56	Learning Arithmetic Circuits	addition affecting after algebraic algorithmic already also analogous ancestor appendix apply applying approach approx approximate arbitrarily arcs arithmetic arnborg artif associative assumed assuming base bayes bayesian been belief between blake boutilier brodley call calls cambridge carlo case chain chapman chechetka chickering child children circuit circuits complex complexity computes conclude conditional conditioned conjunction conjunctions constant construct context copy corneil corresponding darwiche databases decision decomposable defined demonstrate dept deterministic different differential directly discrete discussion disjunction distributions domingos each eachmovie easy efficient embeddings equivalent estimation every evidence exactly except exclusive explorations expressed factoring false fast figure finding found frasca friedman from geiger gibbs gilks goldszmidt graphs greedy guaranteed guestrin hall hardness have having heckerman homes html http hulten icml identical image images independence indicator induction inductively inference initial intel intelligent intl involves irvine iteration jaeger jordan junction kaufmann know known kohavi koller lactr large last leaf learn learnac learning legend lemma likelihood linking links local logical logically lowd mach machine make marginal markov mason master maximum means medium meek meila merz methods microsoft mining mixtures mlearn mlrepository model models monte morgan msweb multiplication must mutually naive negation network networks nielsen nips node nodes number onion order organizers other over parameters pearl peeling performed polynomial practice preserved principled probabilistic probability proc product proof properties proskurowski proved query quick rather reasoning redmond refer references refers remove replace replaced replacing report repository rept research respectively resulting richardson roth sampling satisfied satisfies scenarios seattle second show shown siam sigkdd silander simplifications simplify since sketch slow smooth smoothness specific spiegelhalter split splitac splits srebro straightforward structure such sufficient systems taking tech terms than that then theorem therefore these thesis thin this those time toolkit transformation tree trees true truth univ using value values variable variables verify verifying very washington whenever winmine with without working zheng
http://uai2008.cs.helsinki.fi/UAI_camera_ready/eberhardt.pdf	43	Almost Optimal Intervention Sets for Causal Discovery	acknowledgements active actual acyclic adaptive algorithm algorithms among annual applied approach approximation artificial auai bacchus background based bayes bayesian berkeley bounds brussels cannot carnegie case causal causality causation characterize chickering classes clique coloring complete computation computer concise conclusion conference conjecture consequently constraints cooper corvallis dags data decision department determines discover discovery eberhardt edge edition efficient eleventh equivalence every experiments explanation fellowship folkman foundation francisco from function funded gaussian generalized generate glymour goemans graphs grateful greedy have hoyer hypotheses hypothesis identification identify improved inference intel intelligence international intervention jaakkola james joint journal kaufmann kera knowledge koller largest learning leray ligence linear machine manderick markov mathematics maximum mcdonnell meek meganck mellon minen mixture model monochromatic montreal more morgan much murphy necessary networks number observational oleg optimal optinter oregon oxford passive pearl pikhurko pittsburgh pointing practice prediction press problems procedure procedures proceedings programming provide provided qualitative references relations relevant report representation represented research result resulting rinen satisfiability satisfy scheines science scientific search semidefinite sequences sets shimizu shown siam simulations space spaces spirtes structural structure structures subgraphs sufficient technical than that theorem theoretic therefore thesis think this tong true uncertainty underdetermination undirected universiteit university using variables very vrije williamson with worst
http://uai2008.cs.helsinki.fi/UAI_camera_ready/dawid.pdf	17	Identifying Optimal Sequential Decisions	abbreviations above acyclic adding addison addressed advisable affected after again algorithm also alternative analysis ancestor ancestral anticoagulation appear application applications applying approach approaches argue argued argument artificial assume assumes assumption assumptions auai avoiding barndorff based because been before being berkane between biometrics biometrika biostatistics both cambridge cannot case causal causality certain chapman check child college common complete complex compute computed concatenation condition conditional conditioned conditions conference connected connects consequences consider constructed construction contains context contradiction control covariates cowell create criterion current dags data dawid decision decisions deduce define defined definition demonstrated demystifying denote denotes department descendants descending desired determined diagram diagrams didelez direct directed directly discussion distribution distributions does dynamic earlier effect effects elsevier empirical empty endo estimable estimate estimating estimation evaluation even events every exactly exist exists exit expectation expert exposure expressiveness extended extends fact factors fail fails first fixes following follows formula framework from fullwood fully geneletti general give given graph graphical graphs hall have having heagerty healthy hence henderson here hidden hold holds however hypotheses identifiability identifiable identification identified identify identifying implies important independence independently indirect individual induction inference influence initial input intelligence intermediate internal intervention interventional interventions introduce involved joining joint jrssb just kaufmann kluppelberg latent lauritzen lecture lemma lemmas like ling link links london longitudinal mani markovian mathematical medicine model modeling modelling models modify moodie moral moralisation moralised more morgan mortality motivated murphy must nature need nested networks nielsen node nodes notation note notes observational only optimal original other outputs particular path pearl periods plans point points positive possess practice press probabilistic probability problem proc proceeding proof properties property proves provides question raiffa reasoning recursion recursive references regarding regime regimes regretbased relation remains removing renamed reparameterizing repeatedly replace report research resp result richardson robins rostho same satisfying science seattle section seen semantics semi separation sequel sequence sequential shall show showing shpitser similar simple since single some sound special specialised spiegelhalter springer stability starting statist statistical statistics stephens stewart still stochastic straightforward strategies strategy stripped structural studies subgraph succeeds such suggested suppose survivor sustained symposium systems takes target that then theorem theory there these they this though thus tian together treatment treatments uncertainty unconditional under university using valid variable variables verify verma version violate violating wasserman wesley what when whence where which while whole whose will wish with within worker would write
http://uai2008.cs.helsinki.fi/UAI_camera_ready/laber.pdf	54	Small Sample Inference for Generalization Error in Classification Using the CUD Bound	abalone absolute accuracy addition additive algorithm amer analysis approximated approximation assoc assumption assumptions average bartlett based better bolstered bootstrap bound bounds braga catalog chapman choice classification classifier classifiers combining comparing competitors complexities complexity computationally conclusions confidence conjecture conservative construct constructed constructing contrast control convexity correct coverage cross daniel data david delay demonstrated depend derived deviation diabetes diameter distribution distributional done donut dougherty draws edgeworth edward efficient efron endix erimental error estimating estimation estimators exceeding exhibited expansion experiments feedback figure first flow four from future gaussian generalization guarantees hall hand having heart highlighted hirschberg icml ijcai improvement infocom intel international intervals introduced introduction ionosphere jordan kang kent known kohavi kurtosis langford learn learning level ligence liver mach machine magic maintaining mammogram martin mcauliffe meeting mendelson method methods minimal model more moreover most much multicast need neto normal normality orange pages paper parametric particular pattern perform peter practical prediction preliminary provide provided rademacher rate rates recognition references regardless remains report research results review risk robust rosa rule sample samples schiavo selection sensitive series sets setting shin shows signaling significance simulated sinica small smallest some space spam springer stable statis statistica statistical statistics strongest structural study superior synchronization table target technical test tests that theoretical theory this those tibshirani train training transactions tutorial ulisses under unknown used using validation value weiss well when which while will with work yang years yellow york yuhong zhang
http://uai2008.cs.helsinki.fi/UAI_camera_ready/wang.pdf	68	Continuous Time Dynamic Topic Mo dels	about absolute acad achieves algorithm allocation almost always american amse analysis applying approach april archambeau around assigning author authors average averaged barack based baseline basic bayesian berkeley best better bioinf blei breitling brownian buntine campaign campbell candidates categories cdna cdtm changing chapman civil clinton collapsed combining comp comparable comparison conference continuous cornford corrada corresponding croft cvpr data decomposition deerwester defined delegate democratic dept diffusion dirichlet discovery discrete document documents dumais durbin dynamic election emmanuel engineering error examples expectation exponential families faster figure filtering finding first flat florida forecasting from fundraising furnas general ghahramani girolami graphical griffiths hall harman harrison harshman healthcare hierarchical hillar hofmann hour icml ieee ijcai indexing inference information introduction iowa issues jaakkola jakulin jmlr jordan journal jump kalman koopman kucinich lafferty landauer latent lawler learning linear lines machine markov mccallum meaning methods microarray mixture model models month more motion multiple natl natural networks nevada newman nips nocedal numerical obama opper optimization ornstein others over overview oxford paul perform performance performances performs perona perplexities perplexity perword phys prediction predictive press probabilistic problems proc process processes randomly recount references report retrieval road rogers role rosen sanguinetti saul scene science scientific semantic series sets shawe shen short sigir sigkdd similarly small smyth social society space springer stamp stands state statistics steyvers stochastic straight superdelegate suppl taylor technical text than theory time topic topical topics trans transaction trec trends uhlenbeck univ using variational vegas wainwright wang week welling were west while wisconsin word wright year
http://uai2008.cs.helsinki.fi/UAI_camera_ready/chandrasekaran.pdf	25	Complexity of Inference in Graphical Mo dels	algorithmic algorithms artificial bayesian bollob chernyak chertkov complexity computation computational contraction coop cowell dawid demaine discrete elief expert fomin free genus grada graph graphs grid hminor improved india inference intel international isaac jiaghayi journal kawarabayashi kolkata lauritzen ligence loop mathematics mechanics minor models modern networks onential ounded ounds parameterized probabilistic proc references series spiegelhalter springer statistical symposium systems texts theory thilikos uate using verlag wagner
http://uai2008.cs.helsinki.fi/UAI_camera_ready/sutton.pdf	66	Dyna-Style Planning with Linear Function Approximation and Prioritized Sweeping	abstract accelerating acknowledge actor adaptive advances agents alberta algorithms applied approximate approximating approximation architectures artificial athena atkeson authors automated automatic autonomous baird barto based behavior bertsekas boutilier bowling boyan bradtke coarse coding computing conference contributions control convergence cosmin critic data dearden decision degris delyon department difference differences dimitri dyna dynamic early efficient european exact examples factored fast framework function game general generalization geramifard global goldszmidt gordon gorithms gratefully haven hierarchy high hlynka icore ieee incremental information ingenuity integrated intelligence international introduction iteration joint journal jussila kuvayev lagoudakis learned learning least less linear local machine mark markov mcmahan method methods model models moore muller multiagent national natural neural neuro ninth nserc optimization optimizers paduraru parr peng performance peters planning playing point policy predict press prioritization prioritized problems proceedings processes processing program programming reacting real references reinforcement representations research residual results ring schaal schaeffer scheduling science scientific seppi seventh shape sigaud sigma silver singh sixteenth solvers sparse speed square squares stages stochastic structure substantial successful supported sutton sweeping systems technical temporal tenth thesis this time trajectory transactions tsitsiklis twelfth university update using vijayakumar williams wingate with within work workshop wuillemin yale
http://uai2008.cs.helsinki.fi/UAI_camera_ready/hajishirzi.pdf	30	Sampling First Order Logical Particles	aaai abbeel abduction about accuracy achieves acknowledgements acti action actions advantage aesis algorithm algorithms amir anonymous applications apply artificial bacchus bart bayesian blackwellised boutilier briefcase calculus career carlo combining comments compact computational compute conclusions constraint continue continuous costa cussens dagstuhl darpa data dean decision depots directions discretizing discriminative distance distribution domains domingos doucet dynamic earlier edition effectors exact expected figure filtering first fofa forthcoming freitas friedman future generalize getoor given gordon grants graphical hajishirzi halpern helpful hidden higher horn ieee ijcai inference infinite intelligence introduction jordan kaelbling kanazawa kersting kevin kjaerulff knowledge koller learn learning levesque like logarithmic logic logical machine markov matthew mdps methods michael model models monte muggleton murphy nance ncsa networks nips noisy number observations observed order page paper partially particle pasula pedro pfeffer plato pool porteous posterior practice pram presented press price probabilistic probability processes programming programs project propositional qazi query rabiner raedt raiko raoblackwellised real reasoning recognition references reiter relational representation research reviewers richardson russell samples sampling scai scale scheme selected seminar sensors sequence sequential several shirazi situation speech springer state stochastic supported symbolic system takes taskar techniques temporal text than thank that their there thesis this transition tutorial uiuc understanding value variables vogel welling with work workshop would zettlemoyer
http://uai2008.cs.helsinki.fi/UAI_camera_ready/cai.pdf	37	On Identifying Total Effects in the Presence of Latent Variables and Selection bias	acknowlegdement acyclic addition algorithm analysis ancestral annals application applications applied approach artificial based bayesian bias biases biometrika bowden brito calculus cambridge cases causal causality causation chouaniere classical closed clrendon cochran colliderstratification complete computation condition conditions conference confounderenvironmetrics confounders confounding continuous cooper correlated covariance criteria culture data derive diaz directed discovery discussed distributions does edition editors education effect effects epidemiology equation evaluated even expression factor form foundation from fujiwara furthermore gaussian generalized glymour graph graphical greenland grzebyk help hernan hernandez hidden hold huang identifiability identifiable identification identifying independent inference instrumental intel intervention interventional introduced irikura itoh japan japanese john johnson joint journal judge kaiseki kamigori katayama kotz kougyou kurata kuroki latent lauritzen ligence linear markov matrix mazda meek method methodika ministry miyakawa model modeling models moderating multi multiple multivariate national nikkagiren nineteenth normal observational okeru okuno omission order paper pearl plement pointed population prediction presence press pressoxford problem problems proceeding proceedings process proposed provided quantifying reasoning recovering recursive references regression research residuals results richardson robins royal rules scheines science search selection semimarkovian shpitser single society some sons spirtes sports stanghellini statistical statistics structural structure studies supported tahenryou technology that there this tian tokyo total true turkington uncertainty under university unmeasured unobserved using valtorta variable variables variate versus wermuth when where whether whittaker wild wiley with workshop york
http://uai2008.cs.helsinki.fi/UAI_camera_ready/elidan.pdf	11	Convex Point Estimation using Undirected Bayesian Transfer Hierarchies	abstraction allocation analysis baxter bayes bayesian blei bootstrap boyd carlin cartoons caruna chapman classes classification convex cvpr data dirichlet easier efron elidan empirical first from gelman hall heitz hierarchies hierarchy icml images improving information intelligent introduction jordan journal kaufmann koller latent learn learning louis machine mccallum methods mitchell model morgan multiple multitask nips object optimization ormoneit pearl probabilistic reasoning references research rosenfeld rubin sampling segal shape shrinkage stern systems task text than theoretic thing thrun tibshirani vandenberghe
http://uai2008.cs.helsinki.fi/UAI_camera_ready/welling.pdf	69	Hybrid Variational/Gibbs Collapsed Inference in Topic Models	acknowledgements advances affairs algorithm algorithms allocation annual approach approximate artificial bayesian beal bezoekersbeurs bidyuk biometrika blackwellisation blei bsik buntine carbonetto castella celeux century cognitive collaborative collapsed college comp computational computer computing conditional conference convex cutset dechter derived diebolt dirichlet distribution dutch economic estimating extensions field finland following foundation freitas from function gatsby grant grants griffiths heckerman helsinki here hidden hierarchical hojen icis inference information intelligence interactive jordan journal kurihara latent learning lecture lemma london machine markov mcmc mean methods ministry minka mixture models multinomial networks neural neuroscience newman nips notes part probabilistic problem proceedings processes processing project quaterly recursive references report reported representation research robert russell sampling schemes science scott semantic society sorensen springer statis steyvers supported systems teacher technical thesis tutorial uncertainty unit university variational welling with
http://uai2008.cs.helsinki.fi/UAI_camera_ready/decampos.pdf	8	Strategy Selection in Influence Diagrams using Imprecise Probabilities	airpower algorithm analytical antonucci applications approach approx approximation artif artillery assembly assemblyfacility attack based bayesian belief binary body bodyguard broadcasting cambridge campos capture challenge change commander communications community complexity computational concepts conf convex cooper corp couso cozman cplex credal custody cybernetics darwiche davis decision decisions defense deptula destroy destroyrd diagram diagrams dittmer documentation effects evaluating evaluation exact explanations factorization fagiuoli figure force francisco from grand ground group guard howard http hypothesis hypothetical iads ieee ijar ilog imprecise independence inference inferences inferential influence information integer intel interval jensen joint journal junction know language launch lauritzen ledge ligence limited madison management matheson menlo method minneapolis modeling moral morale multilinear nature networks nilsson occupation operat operations optimization park planning policy polytrees poole press principles probabilistic probabilities probability problem problems programming propaganda propagation properties radars rand rdfacility reason reasoning references report representation representing research researcher results risk science second separation series sets shachter simultaneous solving special specification starting stepwisedecomposable storage storagefacility strategic strategies strike superiority surrender survey symp symposium systems technical territory theoretic theories through transactions trees uncertain uncertainty unified updating using valencia variables volume walley warfare with zaffalon zhang
http://uai2008.cs.helsinki.fi/UAI_camera_ready/sontag.pdf	4	Tightening LP Relaxations for MAP using Message Passing	acknowledgements adding advances agreement algorithm algorithms also alternative anal analyzing applied apply approach approximating approximation approximations aritifical asynchronous basic belief beliefs berkeley best better bioinformatics bound boundminimization bounds chain chazelle choice clusters comparison computed conclusion configurations constructing convergent convex correspondence current darpa dechter decreasing dense dept different difficult dual efficient efficiently elidan empirical energies energy estimation evaluation explanations exponential families fellowship figure finally find finding fixing focused foundation frame free freeman general generalized generation globally globerson graduate graphical graphs greedy hard here hong iccv ideas ieee ijcv implications inference information informed ingredients integer intel into israeli iteration iterations iterative jaakkola jmlr joingraph jordan kask kingsford known koller kolmogorov larger learning ligence linear long lozano lprelaxations mach many maps marginal marginals mateescu mcgraw meltzer message messages methods minimization minka models monotonically morphing mplp necessary needs networks neural next objective optimal optimality order outer paper part passing pattern placement polytope positioning practice presented press previously primarily problem problems processing product program programming propagation protein random real reduction references region regions relaxation relaxations remains report research residual revision reweighted santos scharstein schedules scheduling science shimony showed side similar singh size solutions solve solving sontag square squares start statistics stereo structured study such supported systems szeliski taxonomy technical that theory this thought through tractable trans transfer tree trees triplet used using value variational vision wabi wainwright warm weiss well welling were while willsky with work works world yanover yedidia yields
http://uai2008.cs.helsinki.fi/UAI_camera_ready/tian.pdf	16	Identifying Dynamic Sequential Plans	aaai algorithm annual artifical artificial barndorff bayesian boca causal chapman chapter college complete complex conference consequences dawid decision department diagrams didelez direct dynamic editors effects first foundations geneletti graphical hall heckerman huang identifiability identifying indirect inference influence intel international journal july kluppelberg lauritzen ligence london menlo modelling models national networks nielsen pages park press proceedings raton reasoning references report research review science sequential shachter sound statistical stochastic strategies systems technical theoretic treatment treatments twenty uncertainty university valtorta
http://uai2008.cs.helsinki.fi/UAI_camera_ready/gogate.pdf	44	AND/OR Importance Sampling	achieved adaptive advances aistats algorithm algorithms annual approximate artif artificial based bayesian because bidyuk blackwell blackwellised cambridge carlo cheng combining computations computer conference consistent constraints cutset dawid dechter dependence dictated discrete does druzdzel econometric econometrica editors empirical estimation evidential exact figure fishelson geiger genetic geweke gibbs gogate graph graphical graphs hernndez however hugs hybrid imations importance inference information instance integration intell intelligence international ipmu jair john journal junction kjaerulff large lauritzen linkage mateescu mathematical messages method mixing modelling models monte moral networks neural optimizing pages paskin performance press principles proceedings processing propagation reasoning recent recomb reduction references require rubinstein rulff sample samples samplesearch sampling saul scheme schemes scholkopf search searches seconds similar simulation some sons spaces study such systems that theorem these thrun time tree trees uncertainty using variance while wiley with work york yuan
http://uai2008.cs.helsinki.fi/UAI_camera_ready/brunskill.pdf	12	CORL: A Continuous-state Offset-dynamics Reinforcement Learner	abbeel algorithm apprenticeship automatic brafman brunskill chow conference continuous control corl discrete dynamics exploration general ieee institute international journal learner learning leffler littman machine massachusetts multigrid near offset optimal polynomial proceedings references reinforcement report research state stochastic technical technology tennenholtz time transactions tsitsiklis
http://uai2008.cs.helsinki.fi/UAI_camera_ready/barber.pdf	23	Clique Matrices for Statistical Graph Decomp osition and Parameterising Restricted Positive Definite Matrices	advances algebra algorithm american analysis anderson applications approximate artificial barber bayesian belief biometrika bollen brockington camouflaging chal chaudhuri cliques coloring covariance culberson diestel dimacs distribution drton dunsen edition editor equation estimation fields gaussian graph graphical graphs implementation independence independent inference information intel interscience introduction johnson layered lenge ligence likelihood linear marginal mathematical matrix maximum maximumlikelihood modeling models multivariate networks neural nips normal olkin palomo parameters press processing quasi random references richardon richarson samsi satisfiability second sets sigmoid society sollich springer statistical structural systems theory uncertainty volume wiley with zeros