Lifted Probabilistic Inference for Asymmetric Graphical Models
This work addresses a bottleneck in probabilistic inference for asymmetric models, offering an unbiased solution that improves accuracy for applications in relational data and evidence-based systems, though it is incremental as it builds on existing lifted inference methods.
The paper tackles the problem of performing probabilistic inference in asymmetric graphical models, which are common in real-world scenarios, by introducing a sampling-based framework that uses approximate symmetries within a Metropolis-Hastings Markov chain to produce unbiased probability estimates, with experiments showing it outperforms existing MCMC algorithms.
Lifted probabilistic inference algorithms have been successfully applied to a large number of symmetric graphical models. Unfortunately, the majority of real-world graphical models is asymmetric. This is even the case for relational representations when evidence is given. Therefore, more recent work in the community moved to making the models symmetric and then applying existing lifted inference algorithms. However, this approach has two shortcomings. First, all existing over-symmetric approximations require a relational representation such as Markov logic networks. Second, the induced symmetries often change the distribution significantly, making the computed probabilities highly biased. We present a framework for probabilistic sampling-based inference that only uses the induced approximate symmetries to propose steps in a Metropolis-Hastings style Markov chain. The framework, therefore, leads to improved probability estimates while remaining unbiased. Experiments demonstrate that the approach outperforms existing MCMC algorithms.