lifted inference
On the Complexity and Approximation of Binary Evidence in Lifted Inference
Lifted inference algorithms exploit symmetries in probabilistic models to speed up inference. They show impressive performance when calculating unconditional probabilities in relational models, but often resort to non-lifted inference when computing conditional probabilities. The reason is that conditioning on evidence breaks many of the model's symmetries, which preempts standard lifting techniques. Recent theoretical results show, for example, that conditioning on evidence which corresponds to binary relations is #P-hard, suggesting that no lifting is to be expected in the worst case. In this paper, we balance this grim result by identifying the Boolean rank of the evidence as a key parameter for characterizing the complexity of conditioning in lifted inference. In particular, we show that conditioning on binary evidence with bounded Boolean rank is efficient. This opens up the possibility of approximating evidence by a low-rank Boolean matrix factorization, which we investigate both theoretically and empirically.
Lifted Inference Seen from the Other Side : The Tractable Features
Lifted inference algorithms for representations that combine first-order logic and probabilistic graphical models have been the focus of much recent research. All lifted algorithms developed to date are based on the same underlying idea: take a standard probabilistic inference algorithm (e.g., variable elimination, belief propagation etc.) and improve its efficiency by exploiting repeated structure in the first-order model. In this paper, we propose an approach from the other side in that we use techniques from logic for probabilistic inference. In particular, we define a set of rules that look only at the logical representation to identify models for which exact efficient inference is possible. We show that our rules yield several new tractable classes that cannot be solved efficiently by any of the existing techniques.
Lifted Inference Seen from the Other Side : The Tractable Features
Jha, Abhay, Gogate, Vibhav, Meliou, Alexandra, Suciu, Dan
Lifted inference algorithms for representations that combine first-order logic and probabilistic graphical models have been the focus of much recent research. All lifted algorithms developed to date are based on the same underlying idea: take a standard probabilistic inference algorithm (e.g., variable elimination, belief propagation etc.) and improve its efficiency by exploiting repeated structure in the first-order model. In this paper, we propose an approach from the other side in that we use techniques from logic for probabilistic inference. In particular, we define a set of rules that look only at the logical representation to identify models for which exact efficient inference is possible. We show that our rules yield several new tractable classes that cannot be solved efficiently by any of the existing techniques.
On the Complexity and Approximation of Binary Evidence in Lifted Inference
Broeck, Guy Van den, Darwiche, Adnan
Lifted inference algorithms exploit symmetries in probabilistic models to speed up inference. They show impressive performance when calculating unconditional probabilities in relational models, but often resort to non-lifted inference when computing conditional probabilities. The reason is that conditioning on evidence breaks many of the model's symmetries, which preempts standard lifting techniques. Recent theoretical results show, for example, that conditioning on evidence which corresponds to binary relations is #P-hard, suggesting that no lifting is to be expected in the worst case. In this paper, we balance this grim result by identifying the Boolean rank of the evidence as a key parameter for characterizing the complexity of conditioning in lifted inference.
Towards High-Level Probabilistic Reasoning with Lifted Inference
Broeck, Guy Van den (KU Leuven)
High-level representations of uncertainty, such as probabilistic logics and programs, have been around for decades. Lifted inference was initially motivated by the need to make reasoning algorithms high-level as well. While the lifted inference community focused on machine learning applications, the high-level reasoning goal has received less attention recently. We revisit the idea and look at the capabilities of the latest techniques in lifted inference. This lets us conclude that lifted inference is strictly more powerful than propositional inference on high-level reasoning tasks.
Understanding the Complexity of Lifted Inference and Asymmetric Weighted Model Counting
Gribkoff, Eric (University of Washington) | Broeck, Guy Van den (UCLA, KU Leuven) | Suciu, Dan (University of Washington)
We highlight our work on lifted inference for the asymmetric Weighted First-Order Model Counting problem (WFOMC), which counts the assignments that satisfy a given sentence in first-order logic. This work is at the intersection of probabilistic databases and statistical relational learning. First, we discuss how adding negation can lower the query complexity, and describe the essential element (resolution) necessary to extend a previous algorithm for positive queries to handle queries with negation. Second, we describe our novel dichotomy result for a non-trivial fragment of first-order CNF sentences with negation.Finally, we discuss directions for future work.
On the Complexity and Approximation of Binary Evidence in Lifted Inference
Broeck, Guy Van den (University of California, Los Angeles)
Lifted inference algorithms exploit symmetries in probabilistic models to speed up inference. They show impressive performance when calculating unconditional probabilities in relational models, but often resort to non-lifted inference when computing conditional probabilities, because the evidence breaks many of the model's symmetries.Recent theoretical results paint a grim picture, showing that conditioning on binary relations is #P-hard, and in the worst case, no lifting can be expected. In this paper, we identify Boolean rank of the evidence as a key parameter in the complexity of conditioning. We contrast the hardness result by showing that conditioning on binary evidence with bounded Boolean rank is efficient. This opens up the possibility of approximating evidence by a low-rank Boolean matrix factorization that maintains the model's symmetries and admits efficient lifted inference.
Lifted Inference on Transitive Relations
Pu, Wen (University of Illinois at Urbana-Champaign) | Choi, Jaesik (Lawrence Berkeley Laboratory) | Amir, Eyal (University of Illinois at Urbana-Champaign)
Lifted inference algorithms are able to boost efficiency throughย exploiting symmetries and exchangeability of the underlingย first-order probabilistic models. ย Models with transitive relationsย (e.g., if X and Y are friends and so are Y and Z, then X and Z willย likely be friends) are essential in social network analysis. With ย n elements in a transitive relation model, the computationalย complexity of exact propositional inference is O(2 n ( n -1)/2 ),ย making it intractable for large domains. However, no tractable exactย inference on the transitive relations has been reported on theย transitive relations. In this paper, we report a novel deterministicย approximate lifted inference algorithm, which efficiently solvesย inference problems on the transitive relations without degeneratingย input models. We introduce an alternative graph representation forย first-order probabilistic models with formulas of homogeneousย bivariate predicates. The new representation, which is closelyย related to exponential-family random graph models, leads to anย efficient deterministic approximate lifting algorithm by exploitingย the asymptotic properties of the state space. We perform experimentsย to verify the effectiveness of the proposed algorithm.
Exact Lifted Inference with Distinct Soft Evidence on Every Object
Bui, Hung B. (SRI International) | Huynh, Tuyen N. (SRI International) | Braz, Rodrigo de Salvo (SRI International)
The presence of non-symmetric evidence has been a barrier for the application of lifted inference since the evidence destroys the symmetry of the first-order probabilistic model. In the extreme case, if distinct soft evidence is obtained about each individual object in the domain then, often, all current exact lifted inference methods reduce to traditional inference at the ground level. However, it is of interest to ask whether the symmetry of the model itself before evidence is obtained can be exploited. We present new results showing that this is, in fact, possible. In particular, we show that both exact maximum a posteriori (MAP) and marginal inference can be lifted for the case of distinct soft evidence on a unary Markov Logic predicate. Our methods result in efficient procedures for MAP and marginal inference for a class of graphical models previously thought to be intractable.
Bisimulation-based Approximate Lifted Inference
Sen, Prithviraj, Deshpande, Amol, Getoor, Lise
There has been a great deal of recent interest in methods for performing lifted inference; however, most of this work assumes that the first-order model is given as input to the system. Here, we describe lifted inference algorithms that determine symmetries and automatically lift the probabilistic model to speedup inference. In particular, we describe approximate lifted inference techniques that allow the user to trade off inference accuracy for computational efficiency by using a handful of tunable parameters, while keeping the error bounded. Our algorithms are closely related to the graph-theoretic concept of bisimulation. We report experiments on both synthetic and real data to show that in the presence of symmetries, run-times for inference can be improved significantly, with approximate lifted inference providing orders of magnitude speedup over ground inference.