Kersting, Kristian


Lifted Inference for Convex Quadratic Programs

AAAI Conferences

Symmetry is the essential element of lifted inferencethat has recently demonstrated the possibility to perform very efficient inference in highly-connected, but symmetric probabilistic models. This raises the question, whether this holds for optimization problems in general.Here we show that for a large classof optimization methods this is actually the case.Specifically, we introduce the concept of fractionalsymmetries of convex quadratic programs (QPs),which lie at the heart of many AI and machine learning approaches,and exploit it to lift, i.e., to compress QPs.These lifted QPs can then be tackled with the usual optimization toolbox (off-the-shelf solvers, cutting plane algorithms,stochastic gradients etc.). If the original QP exhibitssymmetry, then the lifted one will generallybe more compact, and hence more efficient to solve.


Learning Continuous-Time Bayesian Networks in Relational Domains: A Non-Parametric Approach

AAAI Conferences

Many real world applications in medicine, biology, communication networks, web mining, and economics, among others, involve modeling and learning structured stochastic processes that evolve over continuous time. Existing approaches, however, have focused on propositional domains only. Without extensive feature engineering, it is difficult-if not impossible-to apply them within relational domains where we may have varying number of objects and relations among them. We therefore develop the first relational representation called Relational Continuous-Time Bayesian Networks (RCTBNs) that can address this challenge. It features a nonparametric learning method that allows for efficiently learning the complex dependencies and their strengths simultaneously from sequence data. Our experimental results demonstrate that RCTBNs can learn as effectively as state-of-the-art approaches for propositional tasks while modeling relational tasks faithfully.


Computer Science on the Move: Inferring Migration Regularities from the Web via Compressed Label Propagation

AAAI Conferences

Many collective human activities have been shown to exhibit universal patterns. However, the possibility of regularities underlying researcher migration in computer science (CS) has barely been explored at global scale. To a large extend, this is due to official and commercial records being restricted, incompatible between countries, and especially not registered across researchers. We overcome these limitations by building our own, transnational, large-scale dataset inferred from publicly available information on the Web. Essentially, we use Label Propagation (LP) to infer missing geo-tags of author-paper-pairs retrieved from online bibliographies. On this dataset, we then find statistical regularities that explain how researchers in CS move from one place to another. However, although vanilla LP is simple and has been remarkably successful, its run time can suffer from unexploited symmetries of the underlying graph. Consequently, we introduce compressed LP (CLP) that exploits these symmetries to reduce the dimensions of the matrix inverted by LP to obtain optimal labeling scores. We prove that CLP reaches identical labeling scores as LP, while often being significantly faster with lower memory usage.


Reports of the AAAI 2014 Conference Workshops

AI Magazine

The AAAI-14 Workshop program was held Sunday and Monday, July 27–28, 2012, at the Québec City Convention Centre in Québec, Canada. The AAAI-14 workshop program included fifteen workshops covering a wide range of topics in artificial intelligence. The titles of the workshops were AI and Robotics; Artificial Intelligence Applied to Assistive Technologies and Smart Environments; Cognitive Computing for Augmented Human Intelligence; Computer Poker and Imperfect Information; Discovery Informatics; Incentives and Trust in Electronic Communities; Intelligent Cinematography and Editing; Machine Learning for Interactive Systems: Bridging the Gap between Perception, Action and Communication; Modern Artificial Intelligence for Health Analytics; Multiagent Interaction without Prior Coordination; Multidisciplinary Workshop on Advances in Preference Handling; Semantic Cities -- Beyond Open Data to Models, Standards and Reasoning; Sequential Decision Making with Big Data; Statistical Relational AI; and The World Wide Web and Public Health Intelligence. This article presents short summaries of those events.


Reports of the AAAI 2014 Conference Workshops

AI Magazine

The AAAI-14 Workshop program was held Sunday and Monday, July 27–28, 2012, at the Québec City Convention Centre in Québec, Canada. Canada. The AAAI-14 workshop program included fifteen workshops covering a wide range of topics in artificial intelligence. The titles of the workshops were AI and Robotics; Artificial Intelligence Applied to Assistive Technologies and Smart Environments; Cognitive Computing for Augmented Human Intelligence; Computer Poker and Imperfect Information; Discovery Informatics; Incentives and Trust in Electronic Communities; Intelligent Cinematography and Editing; Machine Learning for Interactive Systems: Bridging the Gap between Perception, Action and Communication; Modern Artificial Intelligence for Health Analytics; Multiagent Interaction without Prior Coordination; Multidisciplinary Workshop on Advances in Preference Handling; Semantic Cities — Beyond Open Data to Models, Standards and Reasoning; Sequential Decision Making with Big Data; Statistical Relational AI; and The World Wide Web and Public Health Intelligence. This article presents short summaries of those events.


Power Iterated Color Refinement

AAAI Conferences

Color refinement is a basic algorithmic routine for graph isomorphismtesting and has recently been used for computing graph kernels as well as for lifting belief propagation and linear programming. So far, color refinement has been treated as a combinatorial problem. Instead, we treat it as a nonlinear continuous optimization problem and prove thatit implements a conditional gradient optimizer that can be turned into graph clustering approaches using hashing and truncated power iterations. This shows that color refinement is easy to understand in terms of random walks, easy to implement (matrix-matrix/vector multiplications) and readily parallelizable. We support our theoretical results with experiments on real-world graphs with millions of edges.


Relational Logistic Regression

AAAI Conferences

Logistic regression is a commonly used representation for aggregators in Bayesian belief networks when a child has multiple parents. In this paper we consider extending logistic regression to relational models, where we want to model varying populations and interactions among parents. In this paper, we first examine the representational problems caused by population variation. We show how these problems arise even in simple cases with a single parametrized parent, and propose a linear relational logistic regression which we show can represent arbitrary linear (in population size) decision thresholds, whereas the traditional logistic regression cannot. Then we examine representing interactions among the parents of a child node, and representing non-linear dependency on population size. We propose a multi-parent relational logistic regression which can represent interactions among parents and arbitrary polynomial decision thresholds. Finally, we show how other well-known aggregators can be represented using this relational logistic regression.



The AAAI-13 Conference Workshops

AI Magazine

The AAAI-13 Workshop Program, a part of the 27th AAAI Conference on Artificial Intelligence, was held Sunday and Monday, July 14–15, 2013 at the Hyatt Regency Bellevue Hotel in Bellevue, Washington, USA. The program included 12 workshops covering a wide range of topics in artificial intelligence, including Activity Context-Aware System Architectures (WS-13-05); Artificial Intelligence and Robotics Methods in Computational Biology (WS-13-06); Combining Constraint Solving with Mining and Learning (WS-13-07); Computer Poker and Imperfect Information (WS-13-08); Expanding the Boundaries of Health Informatics Using Artificial Intelligence (WS-13-09); Intelligent Robotic Systems (WS-13-10); Intelligent Techniques for Web Personalization and Recommendation (WS-13-11); Learning Rich Representations from Low-Level Sensors (WS-13-12); Plan, Activity, and Intent Recognition (WS-13-13); Space, Time, and Ambient Intelligence (WS-13-14); Trading Agent Design and Analysis (WS-13-15); and Statistical Relational Artificial Intelligence (WS-13-16).


Reduce and Re-Lift: Bootstrapped Lifted Likelihood Maximization for MAP

AAAI Conferences

By handling whole sets of indistinguishable objects together, lifted belief propagation approaches have rendered large, previously intractable, probabilistic inference problems quickly solvable. In this paper, we show that Kumar and Zilberstein's likelihood maximization (LM) approach to MAP inference is liftable, too, and actually provides additional structure for optimization. Specifically, it has been recognized that some pseudo marginals may converge quickly, turning intuitively into pseudo evidence. This additional evidence typically changes the structure of the lifted network: it may expand or reduce it. The current lifted network, however, can be viewed as an upper bound on the size of the lifted network required to finish likelihood maximization. Consequently, we re-lift the network only if the pseudo evidence yields a reduced network, which can efficiently be computed on the current lifted network. Our experimental results on Ising models, image segmentation and relational entity resolution demonstrate that this bootstrapped LM via "reduce and re-lift" finds MAP assignments comparable to those found by the original LM approach, but in a fraction of the time.