Kevin B. Korb

Rodney T. O'Donnell, Kevin B. Korb, Lloyd Allison

The two most commonly used criteria for assessing causal model discovery with artificial data are edit-distance and Kullback-Leibler divergence, measured from the true model to the learned model. Both of these metrics maximally reward the true model. However, we argue that they are both insufficiently discriminating in judging the relative merits of false models. Edit distance, for example, fails to distinguish between strong and weak probabilistic dependencies. KL divergence, on the other hand, rewards equally all statistically equivalent models, regardless of their different causal claims. We propose an augmented KL divergence, which we call Causal KL (CKL), which takes into account causal relationships which distinguish between observationally equivalent models. Results are presented for three variants of CKL, showing that Causal KL works well in practice.

Yang Li, Kevin B Korb, Lloyd Allison

Causal discovery automates the learning of causal Bayesian networks from data and has been of active interest from their beginning. With the sourcing of large data sets off the internet, interest in scaling up to very large data sets has grown. One approach to this is to parallelize search using Markov Blanket (MB) discovery as a first step, followed by a process of combining MBs in a global causal model. We develop and explore three new methods of MB discovery using Minimum Message Length (MML) and compare them empirically to the best existing methods, whether developed specifically as MB discovery or as feature selection. Our best MML method is consistently competitive and has some advantageous features.

Yang Li, Lloyd Allison, Kevin Korb

Moral graphs were introduced in the 1980s as an intermediate step when transforming a Bayesian network to a junction tree, on which exact belief propagation can be efficiently done. The moral graph of a Bayesian network can be trivially obtained by connecting non-adjacent parents for each node in the Bayesian network and dropping the direction of each edge. Perhaps because the moralization process looks simple, there has been little attention on the properties of moral graphs and their impact in belief propagation on Bayesian networks. This paper addresses the mistaken claim that it has been previously proved that optimal moral graph triangulation with the constraints of minimum fill-in, treewidth or total states is NP-complete. The problems are in fact NP-complete, but they have not previously been proved. We now prove these.

Yang Li, Kevin Korb, Lloyd Allison

A family of Markov blankets in a faithful Bayesian network satisfies the symmetry and consistency properties. In this paper, we draw a bijection between families of consistent Markov blankets and moral graphs. We define the new concepts of weak recursive simpliciality and perfect elimination kits. We prove that they are equivalent to graph morality. In addition, we prove that morality can be decided in polynomial time for graphs with maximum degree less than $5$, but the problem is NP-complete for graphs with higher maximum degrees.

Xuhui Zhang, Kevin B. Korb, Ann E. Nicholson et al.

The causal discovery of Bayesian networks is an active and important research area, and it is based upon searching the space of causal models for those which can best explain a pattern of probabilistic dependencies shown in the data. However, some of those dependencies are generated by causal structures involving variables which have not been measured, i.e., latent variables. Some such patterns of dependency "reveal" themselves, in that no model based solely upon the observed variables can explain them as well as a model using a latent variable. That is what latent variable discovery is based upon. Here we did a search for finding them systematically, so that they may be applied in latent variable discovery in a more rigorous fashion.

Tim Taylor, Alan Dorin, Kevin Korb

The application of evolution in the digital realm, with the goal of creating artificial intelligence and artificial life, has a history as long as that of the digital computer itself. We illustrate the intertwined history of these ideas, starting with the early theoretical work of John von Neumann and the pioneering experimental work of Nils Aall Barricelli. We argue that evolutionary thinking and artificial life will continue to play an integral role in the future development of the digital world.

Julian R. Neil, Chris S. Wallace, Kevin B. Korb

A major problem for the learning of Bayesian networks (BNs) is the exponential number of parameters needed for conditional probability tables. Recent research reduces this complexity by modeling local structure in the probability tables. We examine the use of log-linear local models. While log-linear models in this context are not new (Whittaker, 1990; Buntine, 1991; Neal, 1992; Heckerman and Meek, 1997), for structure learning they are generally subsumed under a naive Bayes model. We describe an alternative interpretation, and use a Minimum Message Length (MML) (Wallace, 1987) metric for structure learning of networks exhibiting causal independence, which we term first-order networks (FONs). We also investigate local model selection on a node-by-node basis.

Kevin B. Korb, Ann Nicholson, Nathalie Jitnah

Poker is ideal for testing automated reasoning under uncertainty. It introduces uncertainty both by physical randomization and by incomplete information about opponents hands.Another source OF uncertainty IS the limited information available TO construct psychological models OF opponents, their tendencies TO bluff, play conservatively, reveal weakness, etc. AND the relation BETWEEN their hand strengths AND betting behaviour. ALL OF these uncertainties must be assessed accurately AND combined effectively FOR ANY reasonable LEVEL OF skill IN the game TO be achieved, since good decision making IS highly sensitive TO those tasks.We describe our Bayesian Poker Program(BPP), which uses a Bayesian network TO model the programs poker hand, the opponents hand AND the opponents playing behaviour conditioned upon the hand, and betting curves which govern play given a probability of winning. The history of play with opponents is used to improve BPPs understanding OF their behaviour.We compare BPP experimentally WITH : a simple RULE - based system; a program which depends exclusively ON hand probabilities(i.e., without opponent modeling); AND WITH human players.BPP has shown itself TO be an effective player against ALL these opponents, barring the better humans.We also sketch out SOME likely ways OF improving play.