Bayesian Networks: Representation, Variable Elimination
نویسنده
چکیده
We can view a Bayesian network as a compact representation of a joint probability distribution (the computational view). We can also view it as a graphical representation of a set of conditional independence statements (the conditional independence view). As it turns out, the two views are equivalent. (For more details please refer to [1] and [2]). In this class, we will focus exclusively on the computational view. Namely, we will look at a Bayesian network as a polynomial representation of a multi-dimensional (typically 1000s of dimensions) joint probability distribution.
منابع مشابه
On Variable Elimination in Discrete Bayesian Network Inference
We are interested in proving that variable elimination (VE) in discrete Bayesian networks always yields a clearly structured conditional probability table (CPT) rather than a potential as universally stated. A Bayesian network consists of a directed acyclic graph and a corresponding set of CPTs. Based on the conditional independencies holding in the directed acyclic graph, the product of the CP...
متن کاملA Fully Polynomial Time Approximation Scheme for Updating Credal Networks of Bounded Treewidth and Number of Variable States
Credal networks lift the precise probability assumption of Bayesian networks, enabling a richer representation of uncertainty in the form of closed convex sets of probability measures. The increase in expressiveness comes at the expense of higher computational costs. In this paper we present a new algorithm which is an extension of the wellknown variable elimination algorithm for computing post...
متن کاملExploiting sparsity and sharing in probabilistic sensor data models
Probabilistic sensor models defined as dynamic Bayesian networks can possess an inherent sparsity that is not reflected in the structure of the network. Classical inference algorithms like variable elimination and junction tree propagation cannot exploit this sparsity. Also, they do not exploit the opportunities for sharing calculations among different time slices of the model. We show that, us...
متن کاملExploiting Contextual Independence In Probabilistic Inference
Bayesian belief networks have grown to prominence because they provide compact representations for many problems for which probabilistic inference is appropriate, and there are algorithms to exploit this compactness. The next step is to allow compact representations of the conditional probabilities of a variable given its parents. In this paper we present such a representation that exploits con...
متن کاملGeneralizing Variable Elimination in Bayesian Networks
This paper describes a generalized version of the variable elimination algorithm for Bayesian networks. Variable elimination computes the marginal probability for some specified set of variables in a network. The algorithm consists of a single pass through a list of data structures called buckets. The generalization presented here adds a second pass to the algorithm and produces the marginal pr...
متن کامل