@Proceedings{UAI2013, title = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, editor = {Ann Nicholson and Padhraic Smyth}, publisher = {PMLR}, series = {Proceedings of Machine Learning Research}, volume = R11 } @InProceedings{pmlr-vR11-nicholson13a, title = {The 29th Uncertainty in Artificial Intelligence Conference: Preface}, author = {Nicholson, Ann and Smyth, Padhraic}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {1--4}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/nicholson13a/nicholson13a.pdf}, url = {https://proceedings.mlr.press/r11/nicholson13a.html}, abstract = {Uncertainty In Artificial Intelligence Proceedings of the Twenty-Ninth Conference (2013) Edited by Ann Nicholson Padhraic Smyth Uncertainty in Artificial Intelligence Proceedings of the Twenty-Ninth Conference (2013) July 12-14, 2013, Bellevue, Washington, United States Edited by Ann Nicholson, Monash University, Australia Padhraic Smyth, University of California, Irvine, United States General Chairs Nando De Freitas, University of British Columbia, Canada, and Oxford University, United Kingdom Max Chickering, Microsoft Research, Redmond, United States Local Arrangements Chair Marina Meila, University of Washington, United States Sponsored by Microsoft Research, Artificial Intelligence Journal, Amazon.com, Google Inc., Charles River Analytics, Toyota InfoTechnology Center, U.S.A., Inc., IBM Research AUAI Press Corvallis, Oregon Cover design (c) Alice Zheng. Published by AUAI Press for Association for Uncertainty in Artificial Intelligence http://auai.org Editorial Office: P.O. Box 866 Corvallis, Oregon 97339 USA Copyright (c) 2013 by AUAI Press All rights reserved Printed in the United States of America No part of this book may be reproduced, stored in a retrieval system, or transmitted in any form or by any means—electronic, mechanical, photocopying, recording, or other- wise—without the prior written permission of the publisher. ISBN 978-0-9749039-9-6}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-adel13a, title = {Generative Multiple-Instance Learning Models For Quantitative Electromyography}, author = {Adel, Tameem and Urner, Ruth and Smith, Benn and Stashuk, Daniel and Lizotte, Dan}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {5--14}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/adel13a/adel13a.pdf}, url = {https://proceedings.mlr.press/r11/adel13a.html}, abstract = {We present a comprehensive study of the use of generative modeling approaches for Multiple-Instance Learning (MIL) problems. In MIL a learner receives training instances grouped together into bags with labels for the bags only (which might not be correct for the comprised instances). Our work was motivated by the task of facilitating the di- agnosis of neuromuscular disorders using sets of motor unit potential trains (MUPTs) de- tected within a muscle which can be cast as a MIL problem. Our approach leads to a state- of-the-art solution to the problem of muscle classification. By introducing and analyzing generative models for MIL in a general frame- work and examining a variety of model struc- tures and components, our work also serves as a methodological guide to modelling MIL tasks. We evaluate our proposed methods both on MUPT datasets and on the MUSK1 dataset, one of the most widely used bench- marks for MIL.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-ahmad13a, title = {Active Sensing as {B}ayes-Optimal Sequential Decision Making}, author = {Ahmad, Sheeraz and Yu, Angela}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {15--24}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/ahmad13a/ahmad13a.pdf}, url = {https://proceedings.mlr.press/r11/ahmad13a.html}, abstract = {Sensory inference under conditions of uncer- tainty is a major problem in both machine learning and computational neuroscience. An important but poorly understood aspect of sensory processing is the role of active sensing. Here, we present a Bayes-optimal inference and control framework for active sensing, C-DAC (Context-Dependent Active Controller). Unlike previously proposed al- gorithms that optimize abstract statistical objectives such as information maximization (Infomax) [Butko and Movellan, 2010] or one-step look-ahead accuracy [Najemnik and Geisler, 2005], our active sensing model di- rectly minimizes a combination of behavioral costs, such as temporal delay, response error, and sensor repositioning cost. We simulate these algorithms on a simple visual search task to illustrate scenarios in which context- sensitivity is particularly beneficial and op- timization with respect to generic statisti- cal objectives particularly inadequate. Mo- tivated by the geometric properties of the C- DAC policy, we present both parametric and non-parametric approximations, which retain context-sensitivity while significantly reduc- ing computational complexity. These ap- proximations enable us to investigate a more complex search problem involving peripheral vision, and we notice that the performance advantage of C-DAC over generic statistical policies is even more evident in this scenario.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-beame13a, title = {Lower Bounds for Exact Model Counting and Applications in Probabilistic Databases}, author = {Beame, Paul and Li, Jerry and Roy, Sudeepa and Suciu, Dan}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {25--34}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/beame13a/beame13a.pdf}, url = {https://proceedings.mlr.press/r11/beame13a.html}, abstract = {The best current methods for exactly com- puting the number of satisfying assignments, or the satisfying probability, of Boolean for- mulas can be seen, either directly or indi- rectly, as building decision-DNNF (decision decomposable negation normal form) repre- sentations of the input Boolean formulas. Decision-DNNFs are a special case of d- DNNFs where d stands for deterministic. We show that any decision-DNNF can be con- verted into an equivalent FBDD (free binary decision diagram) – also known as a read- once branching program (ROBP or 1-BP) – with only a quasipolynomial increase in rep- resentation size in general, and with only a polynomial increase in size in the special case of monotone k-DNF formulas. Lever- aging known exponential lower bounds for FBDDs, we then obtain similar exponen- tial lower bounds for decision-DNNFs which provide lower bounds for the recent algo- rithms. We also separate the power of decision-DNNFs from d-DNNFs and a gener- alization of decision-DNNFs known as AND- FBDDs. Finally we show how these imply exponential lower bounds for natural prob- lems associated with probabilistic databases.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-bilmes13a, title = {Properties of the Lovász-Bregman Divergence with applications to rank aggregation and clustering}, author = {Bilmes, Jeff and Iyer, Rishabh}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {35--44}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/bilmes13a/bilmes13a.pdf}, url = {https://proceedings.mlr.press/r11/bilmes13a.html}, abstract = {We extend the recently introduced theory of Lovász Bregman (LB) divergences [19] in several ways. We show that they represent a distortion between a “score” and an “ordering”, thus providing a new view of rank aggregation and order based clustering with interesting connections to web ranking. We show how the LB divergences have a number of properties akin to many permutation based metrics, and in fact have as special cases forms very similar to the Kendall-$\tau$ metric. We also show how the LB divergences subsume a number of commonly used ranking measures in information retrieval, like NDCG [22] and AUC [35]. Unlike the traditional permutation based metrics, however, the LB divergence naturally captures a notion of “confidence” in the orderings, thus providing a new represen- tation to applications involving aggregating scores as opposed to just orderings. We show how a number of recently used web ranking models are forms of Lovász Bregman rank aggregation and also observe that a natural form of Mallow’s model using the LB diver- gence has been used as conditional ranking models for the “Learning to Rank” problem.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-boots13a, title = {{H}ilbert Space Embeddings of Predictive State Representations}, author = {Boots, Byron and Gordon, Geoffrey and Gretton, Arthur}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {45--54}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/boots13a/boots13a.pdf}, url = {https://proceedings.mlr.press/r11/boots13a.html}, abstract = {Predictive State Representations (PSRs) are an expressive class of models for controlled stochastic processes. PSRs represent state as a set of predictions of future observable events. Because PSRs are defined entirely in terms of observable data, statistically con- sistent estimates of PSR parameters can be learned efficiently by manipulating moments of observed training data. Most learning al- gorithms for PSRs have assumed that actions and observations are finite with low cardinal- ity. In this paper, we generalize PSRs to in- finite sets of observations and actions, using the recent concept of Hilbert space embed- dings of distributions. The essence is to rep- resent the state as one or more nonparamet- ric conditional embedding operators in a Re- producing Kernel Hilbert Space (RKHS) and leverage recent work in kernel methods to es- timate, predict, and update the representa- tion. We show that these Hilbert space em- beddings of PSRs are able to gracefully han- dle continuous actions and observations, and that our learned models outperform compet- ing system identification algorithms on sev- eral prediction benchmarks.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-bui13a, title = {Automorphism Groups of Graphical Models and Lifted Variational Inference}, author = {Bui, Hung and Huynh, Tuyen and Riedel, Sebasitan}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {55--64}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/bui13a/bui13a.pdf}, url = {https://proceedings.mlr.press/r11/bui13a.html}, abstract = {Using the theory of group action, we first in- troduce the concept of the automorphism group of an exponential family or a graphical model, thus formalizing the general notion of symme- try of a probabilistic model. This automorphism group provides a precise mathematical frame- work for lifted inference in the general exponen- tial family. Its group action partitions the set of random variables and feature functions into equivalent classes (called orbits) having identical marginals and expectations. Then the inference problem is effectively reduced to that of com- puting marginals or expectations for each class, thus avoiding the need to deal with each individ- ual variable or feature. We demonstrate the use- fulness of this general framework in lifting two classes of variational approximation for maxi- mum a posteriori (MAP) inference: local linear programming (LP) relaxation and local LP re- laxation with cycle constraints; the latter yields the first lifted variational inference algorithm that operates on a bound tighter than the local con- straints.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-chatterjee13a, title = {POMDPs under Probabilistic Semantics}, author = {Chatterjee, Krishnendu and Chmel{\'i}k, Martin}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {65--74}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/chatterjee13a/chatterjee13a.pdf}, url = {https://proceedings.mlr.press/r11/chatterjee13a.html}, abstract = {We consider partially observable Markov decision processes (POMDPs) with limit- average payoff, where a reward value in the interval [0, 1] is associated to every transi- tion, and the payoffof an infinite path is the long-run average of the rewards. We con- sider two types of path constraints: (i) quan- titative constraint defines the set of paths where the payoffis at least a given thresh- old $\lambda$1 $\in$(0, 1]; and (ii) qualitative constraint which is a special case of quantitative con- straint with $\lambda$1 = 1. We consider the compu- tation of the almost-sure winning set, where the controller needs to ensure that the path constraint is satisfied with probability 1. Our main results for qualitative path constraint are as follows: (i) the problem of deciding the existence of a finite-memory controller is EXPTIME-complete; and (ii) the problem of deciding the existence of an infinite-memory controller is undecidable. For quantitative path constraint we show that the problem of deciding the existence of a finite-memory controller is undecidable.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-claassen13a, title = {Learning Sparse Causal Models is not {NP}-hard}, author = {Claassen, Tom and Mooij, Joris and Heskes, Tom}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {75--84}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/claassen13a/claassen13a.pdf}, url = {https://proceedings.mlr.press/r11/claassen13a.html}, abstract = {This paper shows that causal model discov- ery is not an NP-hard problem, in the sense that for sparse graphs bounded by node de- gree k the sound and complete causal model can be obtained in worst case order N 2(k+2) independence tests, even when latent vari- ables and selection bias may be present. We present a modification of the well-known FCI algorithm that implements the method for an independence oracle, and suggest improve- ments for sample/real-world data versions. It does not contradict any known hardness re- sults, and does not solve an NP-hard prob- lem: it just proves that sparse causal discov- ery is perhaps more complicated, but not as hard as learning minimal Bayesian networks.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-cussens13a, title = {Advances in {B}ayesian Network Learning using Integer Programming}, author = {Cussens, James and Bartlett, Mark}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {85--94}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/cussens13a/cussens13a.pdf}, url = {https://proceedings.mlr.press/r11/cussens13a.html}, abstract = {We consider the problem of learning Bayesian networks (BNs) from complete discrete data. This problem of discrete optimisation is for- mulated as an integer program (IP). We de- scribe the various steps we have taken to al- low efficient solving of this IP. These are (i) efficient search for cutting planes, (ii) a fast greedy algorithm to find high-scoring (per- haps not optimal) BNs and (iii) tightening the linear relaxation of the IP. After relating this BN learning problem to set covering and the multidimensional 0-1 knapsack problem, we present our empirical results. These show improvements, sometimes dramatic, over ear- lier results.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-ermon13a, title = {Optimization With Parity Constraints: From Binary Codes to Discrete Integration}, author = {Ermon, Stefano and Gomes, Carla and Sabharwal, Ashish and Selman, Bart}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {95--104}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/ermon13a/ermon13a.pdf}, url = {https://proceedings.mlr.press/r11/ermon13a.html}, abstract = {Many probabilistic inference tasks involve summations over exponentially large sets. Recently, it has been shown that these prob- lems can be reduced to solving a polyno- mial number of MAP inference queries for a model augmented with randomly gener- ated parity constraints. By exploiting a con- nection with max-likelihood decoding of bi- nary codes, we show that these optimizations are computationally hard. Inspired by iter- ative message passing decoding algorithms, we propose an Integer Linear Programming (ILP) formulation for the problem, enhanced with new sparsification techniques to improve decoding performance. By solving the ILP through a sequence of LP relaxations, we get both lower and upper bounds on the parti- tion function, which hold with high probabil- ity and are much tighter than those obtained with variational methods.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-fu13a, title = {{B}ethe-{ADMM} for Tree Decomposition based Parallel {MAP} Inference}, author = {Fu, Qiang and Wang, Huahua and Banerjee, Arindam}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {105--114}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/fu13a/fu13a.pdf}, url = {https://proceedings.mlr.press/r11/fu13a.html}, abstract = {We consider the problem of maximum a pos- teriori (MAP) inference in discrete graphical models. We present a parallel MAP infer- ence algorithm called Bethe-ADMM based on two ideas: tree-decomposition of the graph and the alternating direction method of multi- pliers (ADMM). However, unlike the standard ADMM, we use an inexact ADMM augmented with a Bethe-divergence based proximal func- tion, which makes each subproblem in ADMM easy to solve in parallel using the sum-product algorithm. We rigorously prove global conver- gence of Bethe-ADMM. The proposed algorithm is extensively evaluated on both synthetic and real datasets to illustrate its effectiveness. Fur- ther, the parallel Bethe-ADMM is shown to scale almost linearly with increasing number of cores.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-gogate13a, title = {Structured Message Passing}, author = {Gogate, Vibhav and Domingos, Pedro}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {115--124}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/gogate13a/gogate13a.pdf}, url = {https://proceedings.mlr.press/r11/gogate13a.html}, abstract = {In this paper, we present structured message passing (SMP), a unifying framework for ap- proximate inference algorithms that take advan- tage of structured representations such as al- gebraic decision diagrams and sparse hash ta- bles. These representations can yield signifi- cant time and space savings over the conven- tional tabular representation when the message has several identical values (context-specific in- dependence) or zeros (determinism) or both in its range. Therefore, in order to fully exploit the power of structured representations, we propose to artificially introduce context-specific indepen- dence and determinism in the messages. This yields a new class of powerful approximate in- ference algorithms which includes popular algo- rithms such as cluster-graph Belief propagation (BP), expectation propagation and particle BP as special cases. We show that our new algo- rithms introduce several interesting bias-variance trade-offs. We evaluate these trade-offs empir- ically and demonstrate that our new algorithms are more accurate and scalable than state-of-the- art techniques.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-goldsmith13a, title = {Approximation of Lorenz-Optimal Solutions in Multiobjective {M}arkov Decision Processes}, author = {Goldsmith, Judy and Hanna, Josiah and Perny, Patrice and Weng, Paul}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {125--134}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/goldsmith13a/goldsmith13a.pdf}, url = {https://proceedings.mlr.press/r11/goldsmith13a.html}, abstract = {This paper is devoted to fair optimization in Multiobjective Markov Decision Processes (MOMDPs). A MOMDP is an extension of the MDP model for planning under uncer- tainty while trying to optimize several re- ward functions simultaneously. This applies to multiagent problems when rewards define individual utility functions, or in multicrite- ria problems when rewards refer to different features. In this setting, we study the deter- mination of policies leading to Lorenz-non- dominated tradeoffs. Lorenz dominance is a refinement of Pareto dominance that was in- troduced in Social Choice for the measure- ment of inequalities. In this paper, we in- troduce methods to efficiently approximate the sets of Lorenz-non-dominated solutions of infinite-horizon, discounted MOMDPs. The approximations are polynomial-sized subsets of those solutions.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-koyejo13a, title = {Constrained {B}ayesian Inference for Low Rank Multitask Learning}, author = {Koyejo, Oluwasanmi and Ghosh, Joydeep}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {135--144}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/koyejo13a/koyejo13a.pdf}, url = {https://proceedings.mlr.press/r11/koyejo13a.html}, abstract = {We present a novel approach for constrained Bayesian inference. Unlike current methods, our approach does not require convexity of the con- straint set. We reduce the constrained variational inference to a parametric optimization over the feasible set of densities and propose a general recipe for such problems. We apply the proposed constrained Bayesian inference approach to mul- titask learning subject to rank constraints on the weight matrix. Further, constrained parameter estimation is applied to recover the sparse con- ditional independence structure encoded by prior precision matrices. Our approach is motivated by reverse inference for high dimensional func- tional neuroimaging, a domain where the high dimensionality and small number of examples re- quires the use of constraints to ensure meaning- ful and effective models. For this application, we propose a model that jointly learns a weight ma- trix and the prior inverse covariance structure be- tween different tasks. We present experimental validation showing that the proposed approach outperforms strong baseline models in terms of predictive performance and structure recovery.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-malone13a, title = {Evaluating Anytime Algorithms for Learning Optimal {B}ayesian Networks}, author = {Malone, Brandon and Yuan, Changhe}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {145--154}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/malone13a/malone13a.pdf}, url = {https://proceedings.mlr.press/r11/malone13a.html}, abstract = {Exact algorithms for learning Bayesian networks guarantee to find provably optimal networks. However, they may fail in difficult learning tasks due to limited time or memory. In this research we adapt several anytime heuristic search-based algorithms to learn Bayesian networks. These algorithms find high-quality solutions quickly, and continually improve the incumbent solution or prove its optimality before resources are ex- hausted. Empirical results show that the any- time window A* algorithm usually finds higher- quality, often optimal, networks more quickly than other approaches. The results also show that, surprisingly, while generating networks with few parents per variable are structurally simpler, they are harder to learn than complex generating net- works with more parents per variable.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-maua13a, title = {On the Complexity of Strong and Epistemic Credal Networks}, author = {Maua, Denis and de Campos, Cassio and Benavoli, Alessio and Antonucci, Alessandro}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {155--164}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/maua13a/maua13a.pdf}, url = {https://proceedings.mlr.press/r11/maua13a.html}, abstract = {Credal networks are graph-based statistical models whose parameters take values in a set, instead of being sharply specified as in traditional statistical models (e.g., Bayesian networks). The computational complexity of inferences on such models depends on the ir- relevance/independence concept adopted. In this paper, we study inferential complexity under the concepts of epistemic irrelevance and strong independence. We show that in- ferences under strong independence are NP- hard even in trees with ternary variables. We prove that under epistemic irrelevance the polynomial time complexity of inferences in credal trees is not likely to extend to more general models (e.g. singly connected networks). These results clearly distinguish networks that admit efficient inferences and those where inferences are most likely hard, and settle several open questions regarding computational complexity.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-mcinerney13a, title = {Learning Periodic Human Behaviour Models from Sparse Data for Crowdsourcing Aid Delivery in Developing Countries}, author = {McInerney, James and Rogers, Alex and Jennings, NIcholas}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {165--174}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/mcinerney13a/mcinerney13a.pdf}, url = {https://proceedings.mlr.press/r11/mcinerney13a.html}, abstract = {In many developing countries, half the popula- tion lives in rural locations, where access to es- sentials such as school materials, mosquito nets, and medical supplies is restricted. We propose an alternative method of distribution (to stan- dard road delivery) in which the existing mo- bility habits of a local population are leveraged to deliver aid, which raises two technical chal- lenges in the areas optimisation and learning. For optimisation, a standard Markov decision pro- cess applied to this problem is intractable, so we provide an exact formulation that takes advan- tage of the periodicities in human location be- haviour. To learn such behaviour models from sparse data (i.e., cell tower observations), we de- velop a Bayesian model of human mobility. Us- ing real cell tower data of the mobility behaviour of 50,000 individuals in Ivory Coast, we find that our model outperforms the state of the art ap- proaches in mobility prediction by at least 25% (in held-out data likelihood). Furthermore, when incorporating mobility prediction with our MDP approach, we find a 81.3% reduction in total delivery time versus routine planning that min- imises just the number of participants in the so- lution path.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-mooij13a, title = {Cyclic Causal Discovery from Continuous Equilibrium Data}, author = {Mooij, Joris and Heskes, Tom}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {175--183}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/mooij13a/mooij13a.pdf}, url = {https://proceedings.mlr.press/r11/mooij13a.html}, abstract = {We propose a method for learning cyclic causal models from a combination of obser- vational and interventional equilibrium data. Novel aspects of the proposed method are its ability to work with continuous data (without assuming linearity) and to deal with feedback loops. Within the context of biochemical re- actions, we also propose a novel way of mod- eling interventions that modify the activity of compounds instead of their abundance. For computational reasons, we approximate the nonlinear causal mechanisms by (coupled) lo- cal linearizations, one for each experimental condition. We apply the method to recon- struct a cellular signaling network from the flow cytometry data measured by Sachs et al. (2005). We show that our method finds evi- dence in the data for feedback loops and that it gives a more accurate quantitative descrip- tion of the data at comparable model com- plexity.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-niinimaki13a, title = {Treedy: A Heuristic for Counting and Sampling Subsets}, author = {Niinim{\"a}ki, Teppo and Koivisto, Mikko}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {184--192}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/niinimaki13a/niinimaki13a.pdf}, url = {https://proceedings.mlr.press/r11/niinimaki13a.html}, abstract = {Consider a collection of weighted subsets of a ground set N. Given a query subset Q of N, how fast can one (1) find the weighted sum over all subsets of Q, and (2) sample a sub- set of Q proportionally to the weights? We present a tree-based greedy heuristic, Treedy, that for a given positive tolerance d answers such counting and sampling queries to within a guaranteed relative error d and total vari- ation distance d, respectively. Experimen- tal results on artificial instances and in ap- plication to Bayesian structure discovery in Bayesian networks show that approximations yield dramatic savings in running time com- pared to exact computation, and that Treedy typically outperforms a previously proposed sorting-based heuristic.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-pacer13a, title = {Evaluating computational models of explanation using human judgments}, author = {Pacer, Michael and Lombrozo, Tania and Griffiths, Thomas and Williams, Joseph and Chen, Xi}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {193--202}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/pacer13a/pacer13a.pdf}, url = {https://proceedings.mlr.press/r11/pacer13a.html}, abstract = {We evaluate four computational models of ex- planation in Bayesian networks by compar- ing model predictions to human judgments. In two experiments, we present human par- ticipants with causal structures for which the models make divergent predictions and either solicit the best explanation for an observed event (Experiment 1) or have participants rate provided explanations for an observed event (Experiment 2). Across two versions of two causal structures and across both exper- iments, we find that the Causal Explanation Tree and Most Relevant Explanation mod- els provide better fits to human data than either Most Probable Explanation or Expla- nation Tree models. We identify strengths and shortcomings of these models and what they can reveal about human explanation. We conclude by suggesting the value of pur- suing computational and psychological inves- tigations of explanation in parallel.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-shpitser13a, title = {Sparse Nested {M}arkov models with Log-linear Parameters}, author = {Shpitser, Ilya and Evans, Robin and Richardson, Thomas and Robins, James}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {203--212}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/shpitser13a/shpitser13a.pdf}, url = {https://proceedings.mlr.press/r11/shpitser13a.html}, abstract = {Hidden variables are ubiquitous in practi- cal data analysis, and therefore modeling marginal densities and doing inference with the resulting models is an important problem in statistics, machine learning, and causal inference. Recently, a new type of graphi- cal model, called the nested Markov model, was developed which captures equality con- straints found in marginals of directed acyclic graph (DAG) models. Some of these con- straints, such as the so called ‘Verma con- straint’, strictly generalize conditional inde- pendence. To make modeling and inference with nested Markov models practical, it is necessary to limit the number of parameters in the model, while still correctly capturing the constraints in the marginal of a DAG model. Placing such limits is similar in spirit to sparsity methods for undirected graphical models, and regression models. In this paper, we give a log-linear parameterization which allows sparse modeling with nested Markov models. We illustrate the advantages of this parameterization with a simulation study.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-sindhwani13a, title = {Scalable Matrix-valued Kernel Learning for High-dimensional Nonlinear Multivariate Regression and Granger Causality}, author = {Sindhwani, Vikas and Minh, Ha Quang and Lozano, Aurelie}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {213--222}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/sindhwani13a/sindhwani13a.pdf}, url = {https://proceedings.mlr.press/r11/sindhwani13a.html}, abstract = {We propose a general matrix-valued mul- tiple kernel learning framework for high- dimensional nonlinear multivariate regression problems. This framework allows a broad class of mixed norm regularizers, includ- ing those that induce sparsity, to be im- posed on a dictionary of vector-valued Repro- ducing Kernel Hilbert Spaces. We develop a highly scalable and eigendecomposition- free algorithm that orchestrates two inex- act solvers for simultaneously learning both the input and output components of separa- ble matrix-valued kernels. As a key appli- cation enabled by our framework, we show how high-dimensional causal inference tasks can be naturally cast as sparse function esti- mation problems, leading to novel nonlinear extensions of a class of Graphical Granger Causality techniques. Our algorithmic de- velopments and extensive empirical studies are complemented by theoretical analyses in terms of Rademacher generalization bounds.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-srivastava13a, title = {Modeling Documents with Deep {B}oltzmann Machines}, author = {Srivastava, Nitish and Salakhutdinov, Ruslan and Hinton, Geoffrey}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {223--231}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/srivastava13a/srivastava13a.pdf}, url = {https://proceedings.mlr.press/r11/srivastava13a.html}, abstract = {We introduce a type of Deep Boltzmann Ma- chine (DBM) that is suitable for extracting distributed semantic representations from a large unstructured collection of documents. We overcome the apparent difficulty of train- ing a DBM with judicious parameter tying. This enables an efficient pretraining algo- rithm and a state initialization scheme for fast inference. The model can be trained just as efficiently as a standard Restricted Boltzmann Machine. Our experiments show that the model assigns better log probability to unseen data than the Replicated Softmax model. Features extracted from our model outperform LDA, Replicated Softmax, and DocNADE models on document retrieval and document classification tasks.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-vianna13a, title = {Bounded Approximate Symbolic Dynamic Programming for Hybrid MDPs}, author = {Vianna, Luis Gustavo and Sanner, Scott and de Barros, Leliane}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {232--241}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/vianna13a/vianna13a.pdf}, url = {https://proceedings.mlr.press/r11/vianna13a.html}, abstract = {Recent advances in symbolic dynamic pro- gramming (SDP) combined with the ex- tended algebraic decision diagram (XADD) data structure have provided exact solutions for mixed discrete and continuous (hybrid) MDPs with piecewise linear dynamics and continuous actions. Since XADD-based ex- act solutions may grow intractably large for many problems, we propose a bounded er- ror compression technique for XADDs that involves the solution of a constrained bilin- ear saddle point problem. Fortuitously, we show that given the special structure of this problem, it can be expressed as a bilevel lin- ear programming problem and solved to op- timality in finite time via constraint gener- ation, despite having an infinite set of con- straints. This solution permits the use of efficient linear program solvers for XADD compression and enables a novel class of bounded approximate SDP algorithms for hybrid MDPs that empirically offers order-of- magnitude speedups over the exact solution in exchange for a small approximation error.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-weller13a, title = {On {MAP} Inference by {MWSS} on Perfect Graphs}, author = {Weller, Adrian and Jebara, Tony}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {242--251}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/weller13a/weller13a.pdf}, url = {https://proceedings.mlr.press/r11/weller13a.html}, abstract = {Finding the most likely (MAP) configuration of a Markov random field (MRF) is NP-hard in general. A promising, recent technique is to reduce the problem to finding a max- imum weight stable set (MWSS) on a de- rived weighted graph, which if perfect, al- lows inference in polynomial time. We de- rive new results for this approach, including a general decomposition theorem for MRFs of any order and number of labels, extensions of results for binary pairwise models with submodular cost functions to higher order, and an exact characterization of which bi- nary pairwise MRFs can be efficiently solved with this method. This defines the power of the approach on this class of models, im- proves our toolbox and expands the range of tractable models.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-zhao13a, title = {Active Learning with Expert Advice}, author = {ZHAO, Peilin and Hoi, Steven}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {252--261}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/zhao13a/zhao13a.pdf}, url = {https://proceedings.mlr.press/r11/zhao13a.html}, abstract = {Conventional learning with expert advice methods assume a learner is always receiv- ing the outcome (e.g., class labels) of every incoming training instance at the end of each trial. In real applications, acquiring the out- come from oracle can be costly or time con- suming. In this paper, we address a new problem of active learning with expert ad- vice, where the outcome of an instance is dis- closed only when it is requested by the on- line learner. Our goal is to learn an accu- rate prediction model by asking the oracle the number of questions as small as possi- ble. To address this challenge, we propose a framework of active forecasters for online active learning with expert advice, which at- tempts to extend two regular forecasters, i.e., Exponentially Weighted Average Forecaster and Greedy Forecaster, to tackle the task of active learning with expert advice. We prove that the proposed algorithms satisfy the Han- nan consistency under some proper assump- tions, and validate the efficacy of our tech- nique by an extensive set of experiments.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-amizadeh13a, title = {The Bregman Variational Dual-Tree Framework}, author = {Amizadeh, Saeed and Thiesson, Bo and Hauskrecht, Milos}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {262--271}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/amizadeh13a/amizadeh13a.pdf}, url = {https://proceedings.mlr.press/r11/amizadeh13a.html}, abstract = {Graph-based methods provide a powerful tool set for many non-parametric frameworks in Machine Learning. In general, the mem- ory and computational complexity of these methods is quadratic in the number of exam- ples in the data which makes them quickly in- feasible for moderate to large scale datasets. A significant effort to find more efficient so- lutions to the problem has been made in the literature. One of the state-of-the-art methods that has been recently introduced is the Variational Dual-Tree (VDT) frame- work. Despite some of its unique features, VDT is currently restricted only to Euclidean spaces where the Euclidean distance quan- tifies the similarity. In this paper, we ex- tend the VDT framework beyond the Eu- clidean distance to more general Bregman di- vergences that include the Euclidean distance as a special case. By exploiting the properties of the general Bregman divergence, we show how the new framework can maintain all the pivotal features of the VDT framework and yet significantly improve its performance in non-Euclidean domains. We apply the pro- posed framework to different text categoriza- tion problems and demonstrate its benefits over the original VDT.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-bach13a, title = {Hinge-loss {M}arkov Random Fields: Convex Inference for Structured Prediction}, author = {Bach, Stephen and Huang, Bert and London, Ben and Getoor, Lise}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {272--281}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/bach13a/bach13a.pdf}, url = {https://proceedings.mlr.press/r11/bach13a.html}, abstract = {Graphical models for structured domains are powerful tools, but the computational com- plexities of combinatorial prediction spaces can force restrictions on models, or re- quire approximate inference in order to be tractable. Instead of working in a combina- torial space, we use hinge-loss Markov ran- dom fields (HL-MRFs), an expressive class of graphical models with log-concave density functions over continuous variables, which can represent confidences in discrete predic- tions. This paper demonstrates that HL- MRFs are general tools for fast and accu- rate structured prediction. We introduce the first inference algorithm that is both scalable and applicable to the full class of HL-MRFs, and show how to train HL-MRFs with several learning algorithms. Our experiments show that HL-MRFs match or surpass the predic- tive performance of state-of-the-art methods, including discrete models, in four application domains.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-balasubramanian13a, title = {High-dimensional Joint Sparsity Random Effects Model for Multi-task Learning}, author = {Balasubramanian, Krishnakumar and Yu, Kai and Zhang, Tong}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {282--291}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/balasubramanian13a/balasubramanian13a.pdf}, url = {https://proceedings.mlr.press/r11/balasubramanian13a.html}, abstract = {Joint sparsity regularization in multi-task learn- ing has attracted much attention in recent years. The traditional convex formulation employs the group Lasso relaxation to achieve joint sparsity across tasks. Although this approach leads to a simple convex formulation, it suffers from sev- eral issues due to the looseness of the relax- ation. To remedy this problem, we view jointly sparse multi-task learning as a specialized ran- dom effects model, and derive a convex relax- ation approach that involves two steps. The first step learns the covariance matrix of the coef- ficients using a convex formulation which we refer to as sparse covariance coding; the sec- ond step solves a ridge regression problem with a sparse quadratic regularizer based on the co- variance matrix obtained in the first step. It is shown that this approach produces an asymptot- ically optimal quadratic regularizer in the mul- titask learning setting when the number of tasks approaches infinity. Experimental results demon- strate that the convex formulation obtained via the proposed model significantly outperforms group Lasso (and related multi-stage formula- tions).}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-belle13a, title = {Reasoning about Probabilities in Dynamic Systems using Goal Regression}, author = {Belle, Vaishak and Levesque, Hector}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {292--301}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/belle13a/belle13a.pdf}, url = {https://proceedings.mlr.press/r11/belle13a.html}, abstract = {Reasoning about degrees of belief in uncertain dy- namic worlds is fundamental to many applications, such as robotics and planning, where actions mod- ify state properties and sensors provide measurements, both of which are prone to noise. With the exception of limited cases such as Gaussian processes over lin- ear phenomena, belief state evolution can be complex and hard to reason with in a general way. This pa- per proposes a framework with new results that allows the reduction of subjective probabilities after sensing and acting, both in discrete and continuous domains, to questions about the initial state only. We build on an expressive probabilistic first-order logical ac- count by Bacchus, Halpern and Levesque, resulting in a methodology that, in principle, can be coupled with a variety of existing inference solutions.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-bigot13a, title = {Probabilistic Conditional Preference Networks}, author = {Bigot, Damien and Zanuttini, Bruno and Fargier, Helene and Mengin, J{\'e}r{\^o}me}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {302--311}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/bigot13a/bigot13a.pdf}, url = {https://proceedings.mlr.press/r11/bigot13a.html}, abstract = {This paper proposes a “probabilistic” exten- sion of conditional preference networks as a way to compactly represent a probability dis- tributions over preference orderings. It stud- ies the probabilistic counterparts of the main reasoning tasks, namely dominance testing and optimisation from the algorithmical and complexity viewpoints. Efficient algorithms for tree-structured probabilistic CP-nets are given. As a by-product we obtain a linear- time algorithm for dominance testing in stan- dard, tree-structured CP-nets.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-bootkrajang13a, title = {Boosting in the presence of label noise}, author = {Bootkrajang, Jakramate and Kaban, Ata}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {312--321}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/bootkrajang13a/bootkrajang13a.pdf}, url = {https://proceedings.mlr.press/r11/bootkrajang13a.html}, abstract = {Boosting is known to be sensitive to label noise. We studied two approaches to improve AdaBoost’s robustness against labelling er- rors. One is to employ a label-noise robust classifier as a base learner, while the other is to modify the AdaBoost algorithm to be more robust. Empirical evaluation shows that a committee of robust classifiers, al- though converges faster than non label-noise aware AdaBoost, is still susceptible to label noise. However, pairing it with the new ro- bust Boosting algorithm we propose here re- sults in a more resilient algorithm under mis- labelling.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-borboudakis13a, title = {Scoring and Searching over {B}ayesian Networks with Causal and Associative Priors}, author = {Borboudakis, Giorgos and Tsamardinos, Ioannis}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {322--331}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/borboudakis13a/borboudakis13a.pdf}, url = {https://proceedings.mlr.press/r11/borboudakis13a.html}, abstract = {A significant theoretical advantage of search- and-score methods for learning Bayesian Net- works is that they can accept informative prior beliefs for each possible network, thus complementing the data. In this paper, a method is presented for assigning priors based on beliefs on the presence or absence of certain paths in the true network. Such be- liefs correspond to knowledge about the pos- sible causal and associative relations between pairs of variables. This type of knowledge naturally arises from prior experimental and observational data, among others. In addi- tion, a novel search-operator is proposed to take advantage of such prior knowledge. Ex- periments show that, using path beliefs im- proves the learning of the skeleton, as well as the edge directions in the network.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-brenner13a, title = {SparsityBoost: A New Scoring Function for Learning {B}ayesian Network Structure}, author = {Brenner, Eliot and Sontag, David}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {332--341}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/brenner13a/brenner13a.pdf}, url = {https://proceedings.mlr.press/r11/brenner13a.html}, abstract = {We give a new consistent scoring function for structure learning of Bayesian networks. In contrast to traditional approaches to score- based structure learning, such as BDeu or MDL, the complexity penalty that we pro- pose is data-dependent and is given by the probability that a conditional independence test correctly shows that an edge cannot ex- ist. What really distinguishes this new scor- ing function from earlier work is that it has the property of becoming computationally easier to maximize as the amount of data in- creases. We prove a polynomial sample com- plexity result, showing that maximizing this score is guaranteed to correctly learn a struc- ture with no false edges and a distribution close to the generating distribution, when- ever there exists a Bayesian network which is a perfect map for the data generating distri- bution. Although the new score can be used with any search algorithm, we give empirical results showing that it is particularly effec- tive when used together with a linear pro- gramming relaxation approach to Bayesian network structure learning.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-brunskill13a, title = {Sample Complexity of Transfer Reinforcement Learning}, author = {Brunskill, Emma and Li, Lihong}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {342--351}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/brunskill13a/brunskill13a.pdf}, url = {https://proceedings.mlr.press/r11/brunskill13a.html}, abstract = {Transferring knowledge across a sequence of reinforcement-learning tasks is challenging, and has a number of important applications. Though there is encouraging empirical evidence that transfer can improve performance in subsequent reinforcement-learning tasks, there has been very little theoretical analysis. In this paper, we intro- duce a new multi-task algorithm for a sequence of reinforcement-learning tasks when each task is sampled independently from (an unknown) dis- tribution over a finite set of Markov decision pro- cesses whose parameters are initially unknown. For this setting, we prove under certain assump- tions that the per-task sample complexity of ex- ploration is reduced significantly due to trans- fer compared to standard single-task algorithms. Our multi-task algorithm also has the desired characteristic that it is guaranteed not to exhibit negative transfer: in the worst case its per-task sample complexity is comparable to the corre- sponding single-task algorithm.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-chen13a, title = {Parallel {G}aussian Process Regression with Low-Rank Covariance Matrix Approximations}, author = {Chen, Jie and Cao, Nannan and Low, Kian Hsiang and Tan, Colin Keng-Yan and Jaillet, Patrick}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {352--361}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/chen13a/chen13a.pdf}, url = {https://proceedings.mlr.press/r11/chen13a.html}, abstract = {Gaussian processes (GP) are Bayesian non- parametric models that are widely used for prob- abilistic regression. Unfortunately, it cannot scale well with large data nor perform real-time predictions due to its cubic time cost in the data size. This paper presents two parallel GP re- gression methods that exploit low-rank covari- ance matrix approximations for distributing the computational load among parallel machines to achieve time efficiency and scalability. We the- oretically guarantee the predictive performances of our proposed parallel GPs to be equivalent to that of some centralized approximate GP regres- sion methods: The computation of their central- ized counterparts can be distributed among par- allel machines, hence achieving greater time effi- ciency and scalability. We analytically compare the properties of our parallel GPs such as time, space, and communication complexity. Empir- ical evaluation on two real-world datasets in a cluster of 20 computing nodes shows that our parallel GPs are significantly more time-efficient and scalable than their centralized counterparts and exact/full GP while achieving predictive per- formances comparable to full GP.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-cheng13a, title = {Convex Relaxations of Bregman Divergence Clustering}, author = {Cheng, Hao and Zhang, Xinhua and Schuurmans, Dale}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {362--371}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/cheng13a/cheng13a.pdf}, url = {https://proceedings.mlr.press/r11/cheng13a.html}, abstract = {Although many convex relaxations of clustering have been proposed in the past decade, current formulations remain restricted to spherical Gaus- sian or discriminative models and are susceptible to imbalanced clusters. To address these short- comings, we propose a new class of convex re- laxations that can be flexibly applied to more general forms of Bregman divergence clustering. By basing these new formulations on normalized equivalence relations we retain additional control on relaxation quality, which allows improvement in clustering quality. We furthermore develop optimization methods that improve scalability by exploiting recent implicit matrix norm methods. In practice, we find that the new formulations are able to efficiently produce tighter clusterings that improve the accuracy of state of the art methods.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-drougard13a, title = {Qualitative possibilistic Mixed-Observable MDPs}, author = {Drougard, Nicolas and Dubois, Didier and Teichteil-K{\"o}nigsbuch, Florent and Farges, Jean-Loup}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {372--381}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/drougard13a/drougard13a.pdf}, url = {https://proceedings.mlr.press/r11/drougard13a.html}, abstract = {Possibilistic and qualitative POMDPs ($\pi$- POMDPs) are counterparts of POMDPs used to model situations where the agent’s initial belief or observation probabilities are imprecise due to lack of past experiences or insufficient data collection. However, like probabilistic POMDPs, optimally solving $\pi$- POMDPs is intractable: the finite belief state space exponentially grows with the number of system’s states. In this paper, a possibilis- tic version of Mixed-Observable MDPs is pre- sented to get around this issue: the complex- ity of solving $\pi$-POMDPs, some state vari- ables of which are fully observable, can be then dramatically reduced. A value iteration algorithm for this new formulation under in- finite horizon is next proposed and the op- timality of the returned policy (for a spec- ified criterion) is shown assuming the exis- tence of a ”stay” action in some goal states. Experimental work finally shows that this possibilistic model outperforms probabilistic POMDPs commonly used in robotics, for a target recognition problem where the agent’s observations are imprecise.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-feldman13a, title = {Pushing the Envelope of {M}onte-{C}arlo Planning: Formal Guarantees Meet Practical Efficiency}, author = {Feldman, Zohar and Domshlak, Carmel}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {382--391}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/feldman13a/feldman13a.pdf}, url = {https://proceedings.mlr.press/r11/feldman13a.html}, abstract = {Popular Monte-Carlo tree search (MCTS) al- gorithms for online planning, such as $\varepsilon$-greedy tree search and UCT, aim at rapidly iden- tifying a reasonably good action, but pro- vide rather poor worst-case guarantees on performance improvement over time. In con- trast, a recently introduced MCTS algorithm BRUE guarantees exponential-rate improve- ment over time, yet it is not geared towards identifying reasonably good choices right at the go. We take a stand on the individual strengths of these two classes of algorithms, and show how they can be effectively con- nected. We then rationalize a principle of “selective tree expansion”, and suggest a con- crete implementation of this principle within MCTS. The resulting algorithms favorably compete with other MCTS algorithms under short planning times, while preserving the at- tractive convergence properties of BRUE.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-ganti13a, title = {Building Bridges: Viewing Active Learning from the Multi-Armed Bandit Lens}, author = {Ganti, Ravi and Gray, Alexander}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {392--401}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/ganti13a/ganti13a.pdf}, url = {https://proceedings.mlr.press/r11/ganti13a.html}, abstract = {In this paper we propose a multi-armed ban- dit inspired, pool based active learning algo- rithm for the problem of binary classification. By carefully constructing an analogy between active learning and multi-armed bandits, we utilize ideas such as lower confidence bounds, and self-concordant regularization from the multi-armed bandit literature to design our proposed algorithm. Our algorithm is a se- quential algorithm, which in each round as- signs a sampling distribution on the pool, samples one point from this distribution, and queries the oracle for the label of this sam- pled point. The design of this sampling dis- tribution is also inspired by the analogy be- tween active learning and multi-armed ban- dits. We show how to derive lower confidence bounds required by our algorithm. Exper- imental comparisons to previously proposed active learning algorithms show superior per- formance on some standard UCI data-sets.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-geramifard13a, title = {Batch-iFDD for Representation Expansion in Large MDPs}, author = {Geramifard, Alborz and Walsh, Tom and Roy, Nicholas and How, Jonathan}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {402--411}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/geramifard13a/geramifard13a.pdf}, url = {https://proceedings.mlr.press/r11/geramifard13a.html}, abstract = {Matching pursuit (MP) methods are a prom- ising class of feature construction algorithms for value function approximation. Yet exist- ing MP methods require creating a pool of potential features, mandating expert knowl- edge or enumeration of a large feature pool, both of which hinder scalability. This pa- per introduces batch incremental feature de- pendency discovery (Batch-iFDD) as an MP method that inherits a provable convergence property. Additionally, Batch-iFDD does not require a large pool of features, leading to lower computational complexity. Empiri- cal policy evaluation results across three do- mains with up to one million states highlight the scalability of Batch-iFDD over the previ- ous state of the art MP algorithm.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-hajimirsadeghi13a, title = {Multiple Instance Learning by Discriminative Training of {M}arkov Networks}, author = {Hajimirsadeghi, Hossein and Li, jinling and Mori, Greg and Sayed, Tarek and Zaki, Mohammad}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {412--421}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/hajimirsadeghi13a/hajimirsadeghi13a.pdf}, url = {https://proceedings.mlr.press/r11/hajimirsadeghi13a.html}, abstract = {We introduce a graphical framework for multiple instance learning (MIL) based on Markov networks. This framework can be used to model the traditional MIL definition as well as more general MIL definitions. Dif- ferent levels of ambiguity – the portion of positive instances in a bag – can be explored in weakly supervised data. To train these models, we propose a discriminative max- margin learning algorithm leveraging efficient inference for cardinality-based cliques. The efficacy of the proposed framework is evalu- ated on a variety of data sets. Experimental results verify that encoding or learning the degree of ambiguity can improve classifica- tion performance.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-halpern13a, title = {Unsupervised Learning of Noisy-{OR} {B}ayesian Networks}, author = {Halpern, Yonatan and Sontag, David}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {422--431}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/halpern13a/halpern13a.pdf}, url = {https://proceedings.mlr.press/r11/halpern13a.html}, abstract = {This paper considers the problem of learn- ing the parameters in Bayesian networks of discrete variables with known structure and hidden variables. Previous approaches in these settings typically use expectation maximization; when the network has high treewidth, the required expectations might be approximated using Monte Carlo or vari- ational methods. We show how to avoid inference altogether during learning by giv- ing a polynomial-time algorithm based on the method-of-moments, building upon re- cent work on learning discrete-valued mix- ture models. In particular, we show how to learn the parameters for a family of bipartite noisy-or Bayesian networks. In our experi- mental results, we demonstrate an applica- tion of our algorithm to learning QMR-DT, a large Bayesian network used for medical di- agnosis. We show that it is possible to fully learn the parameters of QMR-DT even when only the findings are observed in the training data (ground truth diseases unknown).}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-hensman13a, title = {{G}aussian Processes for Big Data}, author = {Hensman, James and Fusi, Nicolo and Lawrence, Neil}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {432--440}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/hensman13a/hensman13a.pdf}, url = {https://proceedings.mlr.press/r11/hensman13a.html}, abstract = {We introduce stochastic variational inference for Gaussian process models. This enables the application of Gaussian process (GP) models to data sets containing millions of data points. We show how GPs can be vari- ationally decomposed to depend on a set of globally relevant inducing variables which factorize the model in the necessary manner to perform variational inference. Our ap- proach is readily extended to models with non-Gaussian likelihoods and latent variable models based around Gaussian processes. We demonstrate the approach on a simple toy problem and two real world data sets.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-honorio13a, title = {Inverse Covariance Estimation for High-Dimensional Data in Linear Time and Space: Spectral Methods for Riccati and Sparse Models}, author = {Honorio, Jean and Jaakkola, Tommi}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {441--450}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/honorio13a/honorio13a.pdf}, url = {https://proceedings.mlr.press/r11/honorio13a.html}, abstract = {We propose maximum likelihood estimation for learning Gaussian graphical models with a Gaussian (‘2 2) prior on the parameters. This is in contrast to the commonly used Laplace (‘1) prior for encouraging sparseness. We show that our optimization problem leads to a Riccati matrix equation, which has a closed form solution. We propose an efficient al- gorithm that performs a singular value de- composition of the training data. Our algo- rithm is O(NT 2)-time and O(NT)-space for N variables and T samples. Our method is tailored to high-dimensional problems (N $\gg$ T), in which sparseness promoting methods become intractable. Furthermore, instead of obtaining a single solution for a specific reg- ularization parameter, our algorithm finds the whole solution path. We show that the method has logarithmic sample complexity under the spiked covariance model. We also propose sparsification of the dense solution with provable performance guarantees. We provide techniques for using our learnt mod- els, such as removing unimportant variables, computing likelihoods and conditional distri- butions. Finally, we show promising results in several gene expressions datasets.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-hyttinen13a, title = {Discovering Cyclic Causal Models with Latent Variables: A General {SAT}-Based Procedure}, author = {Hyttinen, Antti and Hoyer, Patrik and Eberhardt, Frederick and J{\"a}rvisalo, Matti}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {451--460}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/hyttinen13a/hyttinen13a.pdf}, url = {https://proceedings.mlr.press/r11/hyttinen13a.html}, abstract = {We present a very general approach to learn- ing the structure of causal models based on d-separation constraints, obtained from any given set of overlapping passive observational or experimental data sets. The procedure al- lows for both directed cycles (feedback loops) and the presence of latent variables. Our ap- proach is based on a logical representation of causal pathways, which permits the integra- tion of quite general background knowledge, and inference is performed using a Boolean satisfiability (SAT) solver. The procedure is complete in that it exhausts the available in- formation on whether any given edge can be determined to be present or absent, and re- turns “unknown” otherwise. Many existing constraint-based causal discovery algorithms can be seen as special cases, tailored to cir- cumstances in which one or more restricting assumptions apply. Simulations illustrate the effect of these assumptions on discovery and how the present algorithm scales.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-iwata13a, title = {Warped Mixtures for Nonparametric Cluster Shapes}, author = {Iwata, Tomoharu and Duvenaud, David and Ghahramani, Zoubin}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {461--470}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/iwata13a/iwata13a.pdf}, url = {https://proceedings.mlr.press/r11/iwata13a.html}, abstract = {A mixture of Gaussians fit to a single curved or heavy-tailed cluster will report that the data contains many clusters. To produce more appropriate clusterings, we introduce a model which warps a latent mixture of Gaussians to produce nonparametric cluster shapes. The possibly low-dimensional latent mixture model allows us to summarize the properties of the high-dimensional clusters (or density manifolds) describing the data. The number of manifolds, as well as the shape and dimension of each manifold is automat- ically inferred. We derive a simple inference scheme for this model which analytically inte- grates out both the mixture parameters and the warping function. We show that our model is effective for density estimation, per- forms better than infinite Gaussian mixture models at recovering the true number of clus- ters, and produces interpretable summaries of high-dimensional datasets.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-khaled13a, title = {Solving Limited-Memory Influence Diagrams Using Branch-and-Bound Search}, author = {Khaled, Arindam and Yuan, Changhe and Hansen, Eric}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {471--480}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/khaled13a/khaled13a.pdf}, url = {https://proceedings.mlr.press/r11/khaled13a.html}, abstract = {A limited-memory influence diagram (LIMID) generalizes a traditional influence diagram by relaxing the assumptions of regularity and no- forgetting, allowing a wider range of decision problems to be modeled. Algorithms for solving traditional influence diagrams are not easily gen- eralized to solve LIMIDs, however, and only re- cently have exact algorithms for solving LIMIDs been developed. In this paper, we introduce an exact algorithm for solving LIMIDs that is based on branch-and-bound search. Our approach is re- lated to the approach of solving an influence di- agram by converting it to an equivalent decision tree, with the difference that the LIMID is con- verted to a much smaller decision graph that can be searched more efficiently.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-kumar13a, title = {Collective Diffusion Over Networks: Models and Inference}, author = {Kumar, Akshat and Sheldon, Dan and Srivastava, Biplav}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {481--490}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/kumar13a/kumar13a.pdf}, url = {https://proceedings.mlr.press/r11/kumar13a.html}, abstract = {Diffusion processes in networks are increas- ingly used to model the spread of informa- tion and social influence. In several applica- tions in computational sustainability such as the spread of wildlife, infectious diseases and traffic mobility pattern, the observed data of- ten consists of only aggregate information. In this work, we present new models that gener- alize standard diffusion processes to such col- lective settings. We also present optimization based techniques that can accurately learn the underlying dynamics of the given conta- gion process, including the hidden network structure, by only observing the time a node becomes active and the associated aggregate information. Empirically, our technique is highly robust and accurately learns network structure with more than 90% recall and pre- cision. Results on real-world flu spread data in the US confirm that our technique can also accurately model infectious disease spread.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-langford13a, title = {Normalized Online Learning}, author = {Langford, John and Mineiro, Paul and Ross, Stephane}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {491--499}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/langford13a/langford13a.pdf}, url = {https://proceedings.mlr.press/r11/langford13a.html}, abstract = {We introduce online learning algorithms which are independent of feature scales, proving regret bounds dependent on the ratio of scales existent in the data rather than the absolute scale. This has several useful effects: there is no need to pre- normalize data, the test-time and test-space com- plexity are reduced, and the algorithms are more robust.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-lee13a, title = {Causal Transportability of Experiments on Controllable Subsets of Variables: z-Transportability}, author = {Lee, Sanghack and Honavar, Vasant}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {500--509}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/lee13a/lee13a.pdf}, url = {https://proceedings.mlr.press/r11/lee13a.html}, abstract = {We introduce z-transportability, the problem of estimating the causal effect of a set of vari- ables X on another set of variables Y in a target domain from experiments on any sub- set of controllable variables Z where Z is an arbitrary subset of observable variables V in a source domain. z-Transportability general- izes z-identifiability, the problem of estimat- ing in a given domain the causal effect of X on Y from surrogate experiments on a set of variables Z such that Z is disjoint from X. z- Transportability also generalizes transporta- bility which requires that the causal effect of X on Y in the target domain be estimable from experiments on any subset of all ob- servable variables in the source domain. We first generalize z-identifiability to allow cases where Z is not necessarily disjoint from X. Then, we establish a necessary and sufficient condition for z-transportability in terms of generalized z-identifiability and transporta- bility. We provide a sound and complete al- gorithm that determines whether a causal ef- fect is z-transportable; and if it is, produces a transport formula, that is, a recipe for es- timating the causal effect of X on Y in the target domain using information elicited from the results of experimental manipulations of Z in the source domain and observational data from the target domain. Our results also show that do-calculus is complete for z- transportability.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-maier13a, title = {A Sound and Complete Algorithm for Learning Causal Models from Relational Data}, author = {Maier, Marc and Marazopoulou, Katerina and Arbour, David and Jensen, David}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {510--519}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/maier13a/maier13a.pdf}, url = {https://proceedings.mlr.press/r11/maier13a.html}, abstract = {The PC algorithm learns maximally oriented causal Bayesian networks. However, there is no equivalent complete algorithm for learning the structure of relational models, a more ex- pressive generalization of Bayesian networks. Recent developments in the theory and repre- sentation of relational models support lifted reasoning about conditional independence. This enables a powerful constraint for ori- enting bivariate dependencies and forms the basis of a new algorithm for learning struc- ture. We present the relational causal discov- ery (RCD) algorithm that learns causal rela- tional models. We prove that RCD is sound and complete, and we present empirical re- sults that demonstrate effectiveness.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-meshi13a, title = {Learning Max-margin Tree Predictors}, author = {Meshi, Ofer and Eban, Elad and Elidan, Gal and Globerson, Amir}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {520--529}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/meshi13a/meshi13a.pdf}, url = {https://proceedings.mlr.press/r11/meshi13a.html}, abstract = {Structured prediction is a powerful frame- work for coping with joint prediction of interacting outputs. A central difficulty in using this framework is that often the correct label dependence structure is unknown. At the same time, we would like to avoid an overly complex structure that will lead to intractable prediction. In this work we ad- dress the challenge of learning tree structured predictive models that achieve high accuracy while at the same time facilitate efficient (linear time) inference. We start by proving that this task is in general NP-hard, and then suggest an approximate alternative. Our CRANK approach relies on a novel Circuit- RANK regularizer that penalizes non-tree structures and can be optimized using a convex-concave procedure. We demonstrate the effectiveness of our approach on several domains and show that its accuracy matches that of fully connected models, while per- forming prediction substantially faster.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-mezuman13a, title = {Tighter Linear Program Relaxations for High Order Graphical Models}, author = {Mezuman, Elad and Tarlow, Daniel and Globerson, Amir and Weiss, Yair}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {530--539}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/mezuman13a/mezuman13a.pdf}, url = {https://proceedings.mlr.press/r11/mezuman13a.html}, abstract = {Graphical models with High Order Potentials (HOPs) have received considerable interest in recent years. While there are a variety of ap- proaches to inference in these models, nearly all of them amount to solving a linear pro- gram (LP) relaxation with unary consistency constraints between the HOP and the indi- vidual variables. In many cases, the resulting relaxations are loose, and in these cases the results of inference can be poor. It is thus de- sirable to look for more accurate ways of per- forming inference. In this work, we study the LP relaxations that result from enforcing ad- ditional consistency constraints between the HOP and the rest of the model. We address theoretical questions about the strength of the resulting relaxations compared to the re- laxations that arise in standard approaches, and we develop practical and efficient mes- sage passing algorithms for optimizing the LPs. Empirically, we show that the LPs with additional consistency constraints lead to more accurate inference on some challeng- ing problems that include a combination of low order and high order terms.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-mooij13b, title = {From Ordinary Differential Equations to Structural Causal Models: the deterministic case}, author = {Mooij, Joris and Janzing, Dominik and Schoelkopf, Bernhard}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {540--548}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/mooij13b/mooij13b.pdf}, url = {https://proceedings.mlr.press/r11/mooij13b.html}, abstract = {We show how, and under which conditions, the equilibrium states of a first-order Ordi- nary Differential Equation (ODE) system can be described with a deterministic Structural Causal Model (SCM). Our exposition sheds more light on the concept of causality as ex- pressed within the framework of Structural Causal Models, especially for cyclic models.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-muandet13a, title = {One-Class Support Measure Machines for Group Anomaly Detection}, author = {Muandet, Krikamol and Schoelkopf, Bernhard}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {549--558}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/muandet13a/muandet13a.pdf}, url = {https://proceedings.mlr.press/r11/muandet13a.html}, abstract = {We propose one-class support measure ma- chines (OCSMMs) for group anomaly detec- tion. Unlike traditional anomaly detection, OCSMMs aim at recognizing anomalous ag- gregate behaviors of data points. The OC- SMMs generalize well-known one-class sup- port vector machines (OCSVMs) to a space of probability measures. By formulating the problem as quantile estimation on distribu- tions, we can establish interesting connec- tions to the OCSVMs and variable kernel density estimators (VKDEs) over the input space on which the distributions are defined, bridging the gap between large-margin meth- ods and kernel density estimators. In partic- ular, we show that various types of VKDEs can be considered as solutions to a class of regularization problems studied in this pa- per. Experiments on Sloan Digital Sky Sur- vey dataset and High Energy Particle Physics dataset demonstrate the benefits of the pro- posed framework in real-world applications.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-munos13a, title = {Finite-Time Analysis of Kernelised Contextual Bandits}, author = {Munos, Remi and Valko, Michal and Korda, Nathaniel and Flaounas, Ilias and Cristianini, Nelo}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {559--568}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/munos13a/munos13a.pdf}, url = {https://proceedings.mlr.press/r11/munos13a.html}, abstract = {We tackle the problem of online reward max- imisation over a large finite set of actions de- scribed by their contexts. We focus on the case when the number of actions is too big to sample all of them even once. However we assume that we have access to the similari- ties between actions’ contexts and that the expected reward is an arbitrary linear func- tion of the contexts’ images in the related re- producing kernel Hilbert space (RKHS). We propose KernelUCB, a kernelised UCB algo- rithm, and give a cumulative regret bound through a frequentist analysis. For contex- tual bandits, the related algorithm GP-UCB turns out to be a special case of our algo- rithm, and our finite-time analysis improves the regret bound of GP-UCB for the agnos- tic case, both in the terms of the kernel- dependent quantity and the RKHS norm of the reward function. Moreover, for the linear kernel, our regret bound matches the lower bound for contextual linear bandits.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-nagano13a, title = {Structured Convex Optimization under Submodular Constraints}, author = {Nagano, Kiyohito and Kawahara, Yoshinobu}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {569--578}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/nagano13a/nagano13a.pdf}, url = {https://proceedings.mlr.press/r11/nagano13a.html}, abstract = {A number of discrete and continuous opti- mization problems in machine learning are related to convex minimization problems un- der submodular constraints. In this paper, we deal with a submodular function with a directed graph structure, and we show that a wide range of convex optimization problems under submodular constraints can be solved much more efficiently than general submod- ular optimization methods by a reduction to a maximum flow problem. Furthermore, we give some applications, including sparse op- timization methods, in which the proposed methods are effective. Additionally, we eval- uate the performance of the proposed method through computational experiments.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-niu13a, title = {Stochastic Rank Aggregation}, author = {Niu, Shuzi and Lan, Yanyan and Guo, Jiafeng and Cheng, Xueqi}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {579--588}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/niu13a/niu13a.pdf}, url = {https://proceedings.mlr.press/r11/niu13a.html}, abstract = {This paper addresses the problem of rank aggregation, which aims to find a consensus ranking among multiple ranking inputs. Tra- ditional rank aggregation methods are deter- ministic, and can be categorized into explicit and implicit methods depending on whether rank information is explicitly or implicitly utilized. Surprisingly, experimental results on real data sets show that explicit rank ag- gregation methods would not work as well as implicit methods, although rank information is critical for the task. Our analysis indicates that the major reason might be the unreli- able rank information from incomplete rank- ing inputs. To solve this problem, we propose to incorporate uncertainty into rank aggrega- tion and tackle the problem in both unsuper- vised and supervised scenario. We call this novel framework stochastic rank aggregation (St.Agg for short). Specifically, we introduce a prior distribution on ranks, and transform the ranking functions or objectives in tradi- tional explicit methods to their expectations over this distribution. Our experiments on benchmark data sets show that the proposed St.Agg outperforms the baselines in both un- supervised and supervised scenarios.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-oren13a, title = {Pay or Play}, author = {Oren, Sigal and Schapira, Michael and Tennenholtz, Moshe}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {589--598}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/oren13a/oren13a.pdf}, url = {https://proceedings.mlr.press/r11/oren13a.html}, abstract = {We introduce the class of pay or play games, which captures scenarios in which each deci- sion maker is faced with a choice between two actions: one with a fixed payoffand another with a payoffdependent on others’ selected actions. This is, arguably, the simplest set- ting that models selection among certain and uncertain outcomes in a multi-agent system. We study the properties of equilibria in such games from both a game-theoretic perspec- tive and a computational perspective. Our main positive result establishes the existence of a semi-strong equilibrium in every such game. We show that although simple, pay or play games contain well-studied environ- ments, e.g., vaccination games. We discuss the interesting implications of our results for these environments.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-petrik13a, title = {Solution Methods for Constrained {M}arkov Decision Process with Continuous Probability Modulation}, author = {Petrik, Marek and Subramanian, Dharmashankar and Marecki, Janusz}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {599--607}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/petrik13a/petrik13a.pdf}, url = {https://proceedings.mlr.press/r11/petrik13a.html}, abstract = {We propose solution methods for previously- unsolved constrained MDPs in which actions can continuously modify the transition probabilities within some acceptable sets. While many meth- ods have been proposed to solve regular MDPs with large state sets, there are few practical approaches for solving constrained MDPs with large action sets. In particular, we show that the continuous action sets can be replaced by their extreme points when the rewards are linear in the modulation. We also develop a tractable opti- mization formulation for concave reward func- tions and, surprisingly, also extend it to non- concave reward functions by using their concave envelopes. We evaluate the effectiveness of the approach on the problem of managing delinquen- cies in a portfolio of loans.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-quadrianto13a, title = {The Supervised {IBP}: Neighbourhood Preserving Infinite Latent Feature Models}, author = {Quadrianto, Novi and Sharmanska, Viktoriia and Knowles, David A. and Ghahramani, Zoubin}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {608--617}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/quadrianto13a/quadrianto13a.pdf}, url = {https://proceedings.mlr.press/r11/quadrianto13a.html}, abstract = {We propose a probabilistic model to infer supervised latent variables in the Hamming space from observed data. Our model al- lows simultaneous inference of the number of binary latent variables, and their values. The latent variables preserve neighbourhood structure of the data in a sense that objects in the same semantic concept have similar latent values, and objects in different con- cepts have dissimilar latent values. We for- mulate the supervised infinite latent variable problem based on an intuitive principle of pulling objects together if they are of the same type, and pushing them apart if they are not. We then combine this principle with a flexible Indian Buffet Process prior on the latent variables. We show that the inferred supervised latent variables can be directly used to perform a nearest neighbour search for the purpose of retrieval. We introduce a new application of dynamically extending hash codes, and show how to effectively cou- ple the structure of the hash codes with con- tinuously growing structure of the neighbour- hood preserving infinite latent feature space.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-ruozzi13a, title = {Beyond Log-Supermodularity: Lower Bounds and the {B}ethe Partition Function}, author = {Ruozzi, Nicholas}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {618--627}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/ruozzi13a/ruozzi13a.pdf}, url = {https://proceedings.mlr.press/r11/ruozzi13a.html}, abstract = {A recent result has demonstrated that the Bethe partition function always lower bounds the true partition function of binary, log- supermodular graphical models. We demon- strate that these results can be extended to other interesting classes of graphical models that are not necessarily binary or log-supermodular: the ferromagnetic Potts model with a uniform external field and its generalizations and special classes of weighted graph homomorphism problems.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-sgouritsa13a, title = {Identifying Finite Mixtures of Nonparametric Product Distributions and Causal Inference of Confounders}, author = {Sgouritsa, Eleni and Janzing, Dominik and Peters, Jonas and Sch{\"o}lkopf, Bernhard}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {628--637}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/sgouritsa13a/sgouritsa13a.pdf}, url = {https://proceedings.mlr.press/r11/sgouritsa13a.html}, abstract = {We propose a kernel method to identify finite mixtures of nonparametric product distribu- tions. It is based on a Hilbert space embed- ding of the joint distribution. The rank of the constructed tensor is equal to the num- ber of mixture components. We present an algorithm to recover the components by par- titioning the data points into clusters such that the variables are jointly conditionally in- dependent given the cluster. This method can be used to identify finite confounders.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-shah13a, title = {Determinantal Clustering Processes - A Nonparametric {B}ayesian Approach to Kernel Based Semi-Supervised Clustering}, author = {Shah, Amar and Ghahramani, Zoubin}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {638--647}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/shah13a/shah13a.pdf}, url = {https://proceedings.mlr.press/r11/shah13a.html}, abstract = {Semi-supervised clustering is the task of clus- tering data points into clusters where only a fraction of the points are labelled. The true number of clusters in the data is often un- known and most models require this param- eter as an input. Dirichlet process mixture models are appealing as they can infer the number of clusters from the data. However, these models do not deal with high dimen- sional data well and can encounter difficulties in inference. We present a novel nonparame- teric Bayesian method to cluster data points without the need to prespecify the number of clusters or to model complicated densities from which data points are assumed to be generated from. The key insight is to use determinants of submatrices of a kernel ma- trix as a measure of how close together a set of points are. We explore some theoretical properties of the model and derive a natural Gibbs based algorithm with MCMC hyper- parameter learning. We test the model on various synthetic and real world data sets.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-soufiani13a, title = {Preference Elicitation For General Random Utility Models}, author = {Soufiani, Hossein Azari and Parkes, David and Xia, Lirong}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {648--657}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/soufiani13a/soufiani13a.pdf}, url = {https://proceedings.mlr.press/r11/soufiani13a.html}, abstract = {This paper discusses General Random Utility Models (GRUMs). These are a class of para- metric models that generate partial ranks over alternatives given attributes of agents and alter- natives. We propose two preference elicitation scheme for GRUMs developed from principles in Bayesian experimental design, one for social choice and the other for personalized choice. We couple this with a general Monte-Carlo- Expectation-Maximization (MC-EM) based al- gorithm for MAP inference under GRUMs. We also prove uni-modality of the likelihood func- tions for a class of GRUMs. We examine the performance of various criteria by experimental studies, which show that the proposed elicitation scheme increases the precision of estimation.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-spirtes13a, title = {Calculation of Entailed Rank Constraints in Partially Non-Linear and Cyclic Models}, author = {Spirtes, Peter}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {658--667}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/spirtes13a/spirtes13a.pdf}, url = {https://proceedings.mlr.press/r11/spirtes13a.html}, abstract = {The Trek Separation Theorem (Sullivant et al. 2010) states necessary and sufficient conditions for a linear directed acyclic graphical model to entail for all possible values of its linear coefficients that the rank of various sub-matrices of the covariance matrix is less than or equal to n, for any given n. In this paper, I extend the Trek Separation Theorem in two ways: I prove that the same necessary and sufficient conditions apply even when the generating model is partially non-linear and contains some cycles. This justifies application of constraint-based causal search algorithms to data generated by a wider class of causal models that may contain non-linear and cyclic relations among the latent variables.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-tenzer13a, title = {Speedy Model Selection ({SMS}) for Copula Models}, author = {Tenzer, Yaniv and Elidan, Gal}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {668--677}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/tenzer13a/tenzer13a.pdf}, url = {https://proceedings.mlr.press/r11/tenzer13a.html}, abstract = {We tackle the challenge of efficiently learning the structure of expressive multivariate real- valued densities of copula graphical models. We start by theoretically substantiating the conjecture that for many copula families the magnitude of Spearman’s rank correlation coefficient is monotonic in the expected con- tribution of an edge in network, namely the negative copula entropy. We then build on this theory and suggest a novel Bayesian ap- proach that makes use of a prior over values of Spearman’s rho for learning copula-based models that involve a mix of copula families. We demonstrate the generalization effective- ness of our highly efficient approach on siz- able and varied real-life datasets.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-tossou13a, title = {Probabilistic inverse reinforcement learning in unknown environments}, author = {Tossou, Aristide and Dimitrakakis, Christos}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {678--686}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/tossou13a/tossou13a.pdf}, url = {https://proceedings.mlr.press/r11/tossou13a.html}, abstract = {We consider the problem of learning by demonstration from agents acting in un- known stochastic Markov environments or games. Our aim is to estimate agent prefer- ences in order to construct improved policies for the same task that the agents are trying to solve. To do so, we extend previous prob- abilistic approaches for inverse reinforcement learning in known MDPs to the case of un- known dynamics or opponents. We do this by deriving two simplified probabilistic mod- els of the demonstrator’s policy and utility. For tractability, we use maximum a posteri- ori estimation rather than full Bayesian in- ference. Under a flat prior, this results in a convex optimisation problem. We find that the resulting algorithms are highly compet- itive against a variety of other methods for inverse reinforcement learning that do have knowledge of the dynamics.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-tripp13a, title = {Approximate {K}alman Filter Q-Learning for Continuous State-Space MDPs}, author = {Tripp, Charles and Shachter, Ross}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {687--696}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/tripp13a/tripp13a.pdf}, url = {https://proceedings.mlr.press/r11/tripp13a.html}, abstract = {We seek to learn an effective policy for a Markov Decision Process (MDP) with con- tinuous states via Q-Learning. Given a set of basis functions over state action pairs we search for a corresponding set of linear weights that minimizes the mean Bellman residual. Our algorithm uses a Kalman filter model to estimate those weights and we have developed a simpler approximate Kalman fil- ter model that outperforms the current state of the art projected TD-Learning methods on several standard benchmark problems.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-venugopal13a, title = {Dynamic Blocking and Collapsing for {G}ibbs Sampling}, author = {Venugopal, Deepak and Gogate, Vibhav}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {697--706}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/venugopal13a/venugopal13a.pdf}, url = {https://proceedings.mlr.press/r11/venugopal13a.html}, abstract = {In this paper, we investigate combining block- ing and collapsing – two widely used strategies for improving the accuracy of Gibbs sampling – in the context of probabilistic graphical mod- els (PGMs). We show that combining them is not straight-forward because collapsing (or elim- inating variables) introduces new dependencies in the PGM and in computation-limited settings, this may adversely affect blocking. We there- fore propose a principled approach for tackling this problem. Specifically, we develop two scor- ing functions, one each for blocking and collaps- ing, and formulate the problem of partitioning the variables in the PGM into blocked and collapsed subsets as simultaneously maximizing both scor- ing functions (i.e., a multi-objective optimization problem). We propose a dynamic, greedy algo- rithm for approximately solving this intractable optimization problem. Our dynamic algorithm periodically updates the partitioning into blocked and collapsed variables by leveraging correla- tion statistics gathered from the generated sam- ples and enables rapid mixing by blocking to- gether and collapsing highly correlated variables. We demonstrate experimentally the clear benefit of our dynamic approach: as more samples are drawn, our dynamic approach significantly out- performs static graph-based approaches by an or- der of magnitude in terms of accuracy.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-xie13a, title = {Integrating document clustering and topic modeling}, author = {Xie, Pengtao and Xing, Eric}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {707--716}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/xie13a/xie13a.pdf}, url = {https://proceedings.mlr.press/r11/xie13a.html}, abstract = {Document clustering and topic modeling are two closely related tasks which can mutu- ally benefit each other. Topic modeling can project documents into a topic space which facilitates effective document cluster- ing. Cluster labels discovered by document clustering can be incorporated into topic models to extract local topics specific to each cluster and global topics shared by all clus- ters. In this paper, we propose a multi-grain clustering topic model (MGCTM) which inte- grates document clustering and topic model- ing into a unified framework and jointly per- forms the two tasks to achieve the overall best performance. Our model tightly couples two components: a mixture component used for discovering latent groups in document col- lection and a topic model component used for mining multi-grain topics including local topics specific to each cluster and global top- ics shared across clusters. We employ varia- tional inference to approximate the posterior of hidden variables and learn model param- eters. Experiments on two datasets demon- strate the effectiveness of our model.}, note = {Reissued by PMLR on 04 October 2026.} } @InProceedings{pmlr-vR11-zhang13a, title = {Bennett-type Generalization Bounds: Large-deviation Case and Faster Rate of Convergence}, author = {Zhang, Chao and Ye, Jieping}, booktitle = {Proceedings of the 29th Conference on Uncertainty in Artificial Intelligence}, pages = {717--725}, year = {2013}, editor = {Nicholson, Ann and Smyth, Padhraic}, volume = {R11}, series = {Proceedings of Machine Learning Research}, month = {12--14 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r11/main/assets/zhang13a/zhang13a.pdf}, url = {https://proceedings.mlr.press/r11/zhang13a.html}, abstract = {In this paper, we present the Bennett-type generalization bounds of the learning pro- cess for i.i.d. samples, and then show that the generalization bounds have a faster rate of convergence than the traditional re- sults. In particular, we first develop two types of Bennett-type deviation inequality for the i.i.d. learning process: one pro- vides the generalization bounds based on the uniform entropy number; the other leads to the bounds based on the Rademacher complexity. We then adopt a new method to obtain the alternative expressions of the Bennett-type generalization bounds, which imply that the bounds have a faster rate o(N -1 2 ) of convergence than the traditional results O(N -1 2 ). Additionally, we find that the rate of the bounds will become faster in the large-deviation case, which refers to a sit- uation where the empirical risk is far away from (at least not close to) the expected risk. Finally, we analyze the asymptotical conver- gence of the learning process and compare our analysis with the existing results.}, note = {Reissued by PMLR on 04 October 2026.} }