Please use this identifier to cite or link to this item: http://scholarbank.nus.edu.sg/handle/10635/16430
 Title: Context-sensitive network: A probabilistic context language for adaptive reasoning Authors: ROHIT JOSHI Keywords: probabilistic graphical models; context-centered computing; machine learning; knowledge representation; bayesian networks; adaptive computing Issue Date: 27-Aug-2009 Source: ROHIT JOSHI (2009-08-27). Context-sensitive network: A probabilistic context language for adaptive reasoning. ScholarBank@NUS Repository. Abstract: This thesis considers the problem of capturing situational variations as "contexts" to model information that holds in specific conditions under uncertainty. We introduce a new asymmetric probabilistic graphical language, Context-Sensitive Network, that extends Bayesian network to domains where the variables (or nodes) and their relations (or edges) are functions of the contexts. CSN aims to support scalable and flexible structural adaptation with varying context atttributes and values, while exploiting the graphical properties of an asymmetric representation. A CSN is a directed bipartite graph that represents the product of Conditional Part Factors (CPFs), a new internal representation for a partition of a conditional probability table (CPT) in a specific context. By properly partitioning the CPT of a target variable in a context-dependent manner, we can exploit both local parameter decomposition and graphical structure decomposition. A CSN also forms the basis of a local context modeling scheme that facilitates knowledge acquisition. We describe the theoretical foundations and the practical considerations of the representation, inference, and learning supported by, as well as an empirical evaluation of the proposed language. We demonstrate that multiple, generic contexts, such as those related to the 5 W"s of a situation - who, what, where, which, and when - can be directly incorporated and integrated; the resulting context-specific graphs are much simpler and more efficient to manipulate for inference and learning. Our representation is particularly useful when there are a large number of relevant context attributes, when the context attributes may vary in different conditions, and when all the context values or evidence may not be known \emph{a priori}. We also evaluate the effectiveness of CSN with two case studies involving actual clinical situations and demonstrate that CSN is expressive enough to handle a wide range of problems involving context in real-life applications. URI: http://scholarbank.nus.edu.sg/handle/10635/16430 Appears in Collections: Ph.D Theses (Open)

###### Files in This Item:
File Description SizeFormatAccess SettingsVersion

OPEN

None

#### Page view(s)

218
checked on Dec 11, 2017