Bayesian inference for statistical abduction using Markov chain Monte Carlo
نویسندگان
چکیده
Abduction is one of the basic logical inferences (deduction, induction and abduction) and derives the best explanations for our observation. Statistical abduction attempts to define a probability distribution over explanations and to evaluate them by their probabilities. The framework of statistical abduction is general since many well-known probabilistic models, i.e., BNs, HMMs and PCFGs, are formulated as statistical abduction. Logic-based probabilistic models (LBPMs) have been developed as a way to combine probabilities and logic, and it enables us to perform statistical abduction. However, most of existing LBPMs impose restrictions on explanations (logical formulas) to realize efficient probability computation and learning. To relax those restrictions, we propose two MCMC (Markov chain Monte Carlo) methods for Bayesian inference on LBPMs using binary decision diagrams. The main advantage of our methods over existing methods is that it has no restriction on formulas. In the context of statistical abduction with Bayesian inference, whereas our deterministic knowledge can be described by logical formulas as rules and facts, our non-deterministic knowledge like frequency and preference can be reflected in a prior distribution in Bayesian inference. To illustrate our methods, we first formulate LDA (latent Dirichlet allocation) which is a well-known generative probabilistic model for bag-of-words as a form of statistical abduction, and compare the learning result of our methods with that of an MCMC method called collapsed Gibbs sampling specialized for LDA. We also apply our methods to diagnosis for failure in a logic circuit and evaluate explanations using a posterior distribution approximated by our method. The experiment shows Bayesian inference achieves better predicting accuracy than that of Maximum likelihood estimation.
منابع مشابه
Inference about the Burr Type III Distribution under Type-II Hybrid Censored Data
This paper presents the statistical inference on the parameters of the Burr type III distribution, when the data are Type-II hybrid censored. The maximum likelihood estimators are developed for the unknown parameters using the EM algorithm method. We provided the observed Fisher information matrix using the missing information principle which is useful for constructing the asymptotic confidence...
متن کاملA Disease Outbreak Prediction Model Using Bayesian Inference: A Case of Influenza
Introduction: One major problem in analyzing epidemic data is the lack of data and high dependency among the available data, which is due to the fact that the epidemic process is not directly observable. Methods: One method for epidemic data analysis to estimate the desired epidemic parameters, such as disease transmission rate and recovery rate, is data ...
متن کاملggmcmc: Analysis of MCMC Samples and Bayesian Inference
ggmcmc is an R package for analyzing Markov chain Monte Carlo simulations from Bayesian inference. By using a well known example of hierarchical/multilevel modeling, the article reviews the potential uses and options of the package, ranging from classical convergence tests to caterpillar plots or posterior predictive checks. This R vignette is based on the article published at the Journal of St...
متن کاملSupporting Text
1. Bayesian Statistical Method The Bayesian Motif Clustering (BMC) algorithm proposed in the main article is based on an explicit statistical model that describes the relationship between the observed motifs and the putative regulons (clusters) and a Markov chain Monte Carlo computational method. We describe first the general Bayesian inference procedure and then its detailed implementation for...
متن کاملNew Approaches in 3D Geomechanical Earth Modeling
In this paper two new approaches for building 3D Geomechanical Earth Model (GEM) were introduced. The first method is a hybrid of geostatistical estimators, Bayesian inference, Markov chain and Monte Carlo, which is called Model Based Geostatistics (MBG). It has utilized to achieve more accurate geomechanical model and condition the model and parameters of variogram. The second approach is the ...
متن کامل