Results 1 
6 of
6
Learning Stochastic Logic Programs
, 2000
"... Stochastic Logic Programs (SLPs) have been shown to be a generalisation of Hidden Markov Models (HMMs), stochastic contextfree grammars, and directed Bayes' nets. A stochastic logic program consists of a set of labelled clauses p:C where p is in the interval [0,1] and C is a firstorder range ..."
Abstract

Cited by 1057 (71 self)
 Add to MetaCart
Stochastic Logic Programs (SLPs) have been shown to be a generalisation of Hidden Markov Models (HMMs), stochastic contextfree grammars, and directed Bayes' nets. A stochastic logic program consists of a set of labelled clauses p:C where p is in the interval [0,1] and C is a firstorder rangerestricted definite clause. This paper summarises the syntax, distributional semantics and proof techniques for SLPs and then discusses how a standard Inductive Logic Programming (ILP) system, Progol, has been modied to support learning of SLPs. The resulting system 1) nds an SLP with uniform probability labels on each definition and nearmaximal Bayes posterior probability and then 2) alters the probability labels to further increase the posterior probability. Stage 1) is implemented within CProgol4.5, which differs from previous versions of Progol by allowing userdefined evaluation functions written in Prolog. It is shown that maximising the Bayesian posterior function involves nding SLPs with short derivations of the examples. Search pruning with the Bayesian evaluation function is carried out in the same way as in previous versions of CProgol. The system is demonstrated with worked examples involving the learning of probability distributions over sequences as well as the learning of simple forms of uncertain knowledge.
The Utility of Knowledge in Inductive Learning
, 1992
"... In this paper, we demonstrate how different forms of background knowledge can be integrated with an inductive method for generating constantfree Horn clause rules. Furthermore, we evaluate, both theoretically and empirically, the effect that these types of knowledge have on the cost of learning a r ..."
Abstract

Cited by 145 (22 self)
 Add to MetaCart
In this paper, we demonstrate how different forms of background knowledge can be integrated with an inductive method for generating constantfree Horn clause rules. Furthermore, we evaluate, both theoretically and empirically, the effect that these types of knowledge have on the cost of learning a rule and on the accuracy of a learned rule. Moreover, we demonstrate that a hybrid explanationbased and inductive learning method can advantageously use an approximate domain theory, even when this theory is incorrect and incomplete. 1 Introduction Most existing systems that combine empirical and explanationbased learning severely restrict the complexity of the language for expressing the concept definition. For example, some systems require that the concept definition be expressed in terms of attributevalue pairs (Lebowitz, 1986; Danyluk, 1989). Others effectively restrict the concept definition language to that of propositional calculus, by only allowing unary predicates (Hirsh, 1989;...
Systems for Knowledge Discovery in Databases
 IEEE Transactions On Knowledge And Data Engineering
, 1993
"... The automated discovery of knowledge in databases is becoming increasingly important as the world's wealth of data continues to grow exponentially. Knowledgediscovery systems face challenging problems from realworld databases which tend to be dynamic, incomplete, redundant, noisy, sparse, and very ..."
Abstract

Cited by 94 (8 self)
 Add to MetaCart
The automated discovery of knowledge in databases is becoming increasingly important as the world's wealth of data continues to grow exponentially. Knowledgediscovery systems face challenging problems from realworld databases which tend to be dynamic, incomplete, redundant, noisy, sparse, and very large. This paper addresses these problems and describes some techniques for handling them. A model of an idealized knowledgediscovery system is presented as a reference for studying and designing new systems. This model is used in the comparison of three systems: CoverStory, EXPLORA, and the Knowledge Discovery Workbench. The deficiencies of existing systems relative to the model reveal several open problems for future research.
Learning Structural Decision Trees from Examples
"... STRUCT is a system that learns structural decision trees from positive and negative examples. The algorithm uses a modification of Pagallo and Haussler's FRINGE algorithm to construct new features in a firstorder representation. Experiments compare the effects of different hypothesis evaluation str ..."
Abstract

Cited by 14 (0 self)
 Add to MetaCart
STRUCT is a system that learns structural decision trees from positive and negative examples. The algorithm uses a modification of Pagallo and Haussler's FRINGE algorithm to construct new features in a firstorder representation. Experiments compare the effects of different hypothesis evaluation strategies, domain representation, and feature construction. STRUCT is also compared with Quinlan's FOIL on two domains. The results show that a modified FRINGE algorithm improves accuracy, but that it is sensitive to the distribution of the examples.
On the Effects of Initialising a Neural Network With Prior Knowledge
 In Proceedings of the International Conference on Neural Information Processing (ICONIPâ€™99
, 1999
"... . This paper quantitatively examines the effects of initialising a Rapid Backprop Network (RBP) with prior domain knowledge expressed in the form of propositional rules. The paper first describes the RBP network and then introduces the RULEIN algorithm which encodes propositional rules as the weigh ..."
Abstract

Cited by 1 (0 self)
 Add to MetaCart
. This paper quantitatively examines the effects of initialising a Rapid Backprop Network (RBP) with prior domain knowledge expressed in the form of propositional rules. The paper first describes the RBP network and then introduces the RULEIN algorithm which encodes propositional rules as the weights of the nodes of the RBP network. A selection of datasets is used to compare networks that began learning from tabula rasa with those that were initialised with varying amounts of domain knowledge prior to the commencement of the learning phase. Network performance is compared in terms of time to converge, accuracy at convergence, and network size at convergence. 1. Introduction. Inductive learning methods such as artificial neural networks generally rely on large amounts of training data and require many training cycles to form concepts. A problem frequently encountered by inductive learning methods is insufficient training examples to achieve good generalisation. The use of an approxim...