Probabilistic reasoning in intelligent systems

1988

Subfields: Probabilistic & Bayesian Methods

Citations in the corpus

• Bayesian Structure Learning with Generative Flow Networks (2022, background): “Bayesian networks (Pearl, 1988) are a popular framework of choice for representing uncertainty about the world in multiple scientific domains, including medical diagnosis (Lau-ritzen and Spiegelhalter, 1988; Heckerman and Nathwani, 1992), molecular biology (Friedman, 2004; Sebastiani et al., 2005),…”
• Natural Language Processing (Almost) from Scratch (2011, background): “Separately training a submodel only ar Xi v makes sense when the training data blocks these additional dependency paths (in the sense of d-separation, Pearl, 1988).”
• Fair Inference on Outcomes (2017, background): “Such models generalize independence models on directed acyclic graphs, also known as Bayesian networks (Pearl 1988), to also encode conditional independence statements on counterfactual random variables (Richardson and Robins 2013).”
• From machine learning to machine reasoning (2011, methodology): “…distributions are highly constrained by the algebraic properties of the probability theory: if we know a subset of these conditional distributions, we can apply Bayesian inference to deduct or constrain additional conditional distributions and therefore answer different questions (Pearl, 1988).”
• Active Learning with Statistical Models (1996, background): “Of particular interest is the class of models known as \belief networks" or \Bayesian networks" (Pearl, 1988; Heckerman, Geiger, & Chickering, 1994).”
• Causal inference in statistics: An overview (2009, methodology): “Such testable implications can be read off the diagrams using a graphical criterion known as d-separation (Pearl, 1988).”
• Mean Field Theory for Sigmoid Belief Networks (1996, methodology): “It is worth noting, moreover, that all the above quali(cid:12)cations apply to Markov networks, and that in this domain, mean (cid:12)eld methods are already well-established.”
• Counterfactual Graphical Models for Longitudinal Mediation Analysis With Unobserved Confounding (2012, background): “The indirect effect is defined similarly, expect we now subtract off the direct influence of A on Y from the total effect of setting A to 1: Indirect Effect = E[Y (1)] − E[Y (1, M(0))] (14) It is not difficult to show that these definition reduce to definitions in terms of regression coefficients gi”
• Sum-product networks: A new deep architecture (2011, background): “Graphical models [22] represent distributions compactly as normalized products of factors: P (X = x) = 1 Z ∏”
• Theoretical Impediments to Machine Learning With Seven Sparks from the Causal Revolution (2018, background): “It tells us, for any given pattern of paths in the model, what pattern of dependencies we should expect to find in the data (Pearl, 1988).”
• Strong completeness and faithfulness in Bayesian networks (1995, methodology): “Pearl's claim is supported by the proof that for any di­ rected acyclic graph G there exists distribution for which G is a perfect map (o r, in the language of this paper, the distribution is faithful to G); see Theorem 10 of Pearl1988 and Geiger and Pearl1988b for a proof.”
• A theory of causal learning in children: causal maps and Bayes nets. (2004, background): “…developed in the computer science, philosophy, and statistical literatures over the last 2 decades (Glymour & Cooper, 1999; Kiiveri & Speed, 1982; Pearl, 1988, 2000; Spirtes et al., 1993, 2001) It provides a general, unified representation of causal hypotheses that otherwise take a variety of…”
• 3D Graph Neural Networks for RGBD Semantic Segmentation (2017, background): “One fundamental algorithm to approximate inference in general MRFs/CRFs is loopy belief propagation (BP) [31, 8].”
• Time and sample efficient discovery of Markov blankets and direct causal relations (2003, methodology): “ABN N and a joint distribution J are faithful to one another, i. every conditional independence entailed by the graph of N and the Markov Condition is also present in J [8] (in the terminology of Pearl [7] N is a perfect map of J).”
• Evaluating the Replicability of Significance Tests for Comparing Learning Algorithms (2004, background): “Using stochastic simulation [9], a collection of 1000 training sets with 300 instances each was created.”