RockIt is a maximum a-posteriori (MAP) query engine for statistical relational models. MAP inference in graphical models is an optimization problem which can be compiled to integer linear programs (ILPs).We describe several advances in translating MAP queries to ILP instances and present the novel meta-algorithm cutting plane aggregation (CPA). CPA exploits local context-specific symmetries and bundles up sets of linear constraints. The resulting counting constraints lead to more compact ILPs and make the symmetry of the ground model more explicit to state-of-the-art ILP solvers. Moreover, RockIt parallelizes most parts of the MAP inference pipeline taking advantage of ubiquitous shared-memory multi-core architectures. We report on extensive experiments with Markov logic network (MLN) benchmarks showing that RockIt outperforms the state-of-the-art systems Alchemy, Markov TheBeast, and Tuffy both in terms of efficiency and quality of results.

Statistical relational learning (SRL) augments probabilistic models with relational representations and facilitates reasoning over sets of objects. When learning the probabilistic parameters for SRL models, however, one often resorts to reasoning over individual objects. To address this challenge, we compile a Markov logic network into a compact and efficient first-order data structure and use weighted first-order model counting to exactly optimize the likelihood of the parameters in a lifted manner. By exploiting the relational structure in the model, it is possible to learn more accurate parameters and dramatically improve the run time of the likelihood calculation. This allows us to calculate the exact likelihood for models where previously only approximate inference was feasible. Results on real-world data sets show that this approach learns more accurate models.

We present a robust framework for complex event recognition that is well-suited for integrating information that varies widely in detail and granularity. Consider the scenario of an agent in an instrumented space performing a complex task while describing what he is doing in a natural manner. The system takes in a variety of information, including objects and gestures recognized by RGB-D and descriptions of events extracted from recognized and parsed speech. The system outputs a complete reconstruction of the agent’s plan, explaining actions in terms of more complex activities and filling in unobserved but necessary events. We show how to use Markov Logic (a probabilistic extension to first order logic) to create a theory in which observations can be partial, noisy, and refer to future or temporally ambiguous events; complex events are composed from simpler events in a manner that exposes their structure for inference and learning; and uncertainty is handled in a sound probabilistic manner. We demonstrate the effectiveness of the approach for tracking cooking plans in the presence of noisy and incomplete observations.

Modal Markov Logic for a single agent has previously been proposed as an extension to propositional Markov logic. While the framework allowed reasoning under the principle of maximum entropy for various modal logics, it is not feasible to apply its counting based inference to reason about the beliefs and knowledge of multiple agents due to magnitude of the numbers involved. We propose a modal extension of propositional Markov logicthat avoids this problem by coarsening the state space.The problem stems from the fact that in the single-agent setting, the state space is only doubly exponential in the number of propositions in the domain, but the state space can potentially become infinite in the multi-agent setting. In addition, the proposed framework adds only the overhead of deciding satisfiability for the chosen modal logic on the top of the complexity of exact inference in propositional Markov logic. The proposed framework allows one to find a distribution that matches probabilities of formulas obtained from training data (or provided by an expert). Finally, we show how one can compute lower and upper bounds on probabilities of arbitrary formulas.

RockIt is a maximum a-posteriori (MAP) query engine for statistical relational models. MAP inference in graphical models is an optimization problem which can be compiled to integer linear programs (ILPs). We describe several advances in translating MAP queries to ILP instances and present the novel meta-algorithm cutting plane aggregation (CPA). CPA exploits local context-specific symmetries and bundles up sets of linear constraints. The resulting counting constraints lead to more compact ILPs and make the symmetry of the ground model more explicit to state-of-the-art ILP solvers. Moreover, RockIt parallelizes most parts of the MAP inference pipeline taking advantage of ubiquitous shared-memory multi-core architectures. We report on extensive experiments with Markov logic network (MLN) benchmarks showing that RockIt outperforms the state-of-the-art systems Alchemy, Markov TheBeast, and Tuffy both in terms of efficiency and quality of results. This paper is a short version of a AAAI publication of the same name.

Relational Markov Decision Processes (RMDPs) offer an elegant formalism that combines probabilistic and relational knowledge representations with the decision-theoretic notions of action and utility. In this paper we motivate RMDPs to address a variety of problems in AI, including open world planning, transfer learning, and relational inference. We describe a symbolic dynamic programming approach via the "template method" which addresses the problem of reasoning about exogenous events. We end with a discussion of the challenges involved and some promising future research directions.

We present a robust framework for complex event recognition that is well-suited for integrating information that varies widely in detail and granularity. Consider the scenario of an agent in an instrumented space performing a complex task while describing what he is doing in a natural manner. The system takes in a variety of information, including objects and gestures recognized by RGB-D and descriptions of events extracted from recognized and parsed speech. The system outputs a complete reconstruction of the agent’s plan, explaining actions in terms of more complex activities and filling in unobserved but necessary events. We show how to use Markov Logic (a probabilistic extension to first order logic) to create a theory in which observations can be partial, noisy, and refer to future or temporally ambiguous events; complex events are composed from simpler events in a manner that exposes their structure for inference and learning; and uncertainty is handled in a sound probabilistic manner. We demonstrate the effectiveness of the approach for tracking cooking plans in the presence of noisy and incomplete observations.

While recent Machine Learning (ML) based techniques for activity recognition show great promise, there remain a number of questions with respect to the relative merits of these techniques. To provide a better understanding of the relative strengths of contemporary Activity Recognition methods, in this paper we present a comparative analysis of Hidden Markov Model, Bayesian, and Support Vector Machine based human activity recognition models. The study builds on both pre-existing and newly annotated data which includes interleaved activities. Results demonstrate that while Support Vector Machine based techniques perform well for all data sets considered, simple representations of sensor histories regularly outperform more complex count based models.

In spite of the significant work that has been done todiscover and recognize activities in the smart home re-search, less attention has been paid to predict the futureactivities that the resident is likely to perform. An ac-tivity prediction module can play a major role in designof a smart home. For instance, by taking advantage ofan activity prediction module, a smart home can learncontext-aware rules to prompt individuals to initiate im-portant activities. In this paper, we propose an activityprediction approach using Bayesian networks. We pro-pose a novel two-step inference process to predict thenext activity features and then to predict the next activ-ity label. We also propose an approach to predict thestart time of the next activity which is based on model-ing the relative start time of the predicted activity usinga continuous normal distribution and outlier detection.We evaluate our proposed models using real data col-lected from two smart home apartments.

There are currently no strong methods for planning in a stochastic domain, with low-level sensors that are limited and possibly inaccurate. Existing architectures have flaws that make their use in a real-world environment impractical. We propose an architecture that utilizes POMDPs to create a hierarchical planning system. This system is capable of developing macro-actions that can expedite planning on a large scale, and can learn new plans quickly and efficiently, without deliberate design by the programmer.