Results 1  10
of
10
PALO: A Probabilistic HillClimbing Algorithm
 Artificial Intelligence
, 1995
"... Many learning systems search through a space of possible performance elements, seeking an element whose expected utility, over the distribution of problems, is high. As the task of finding the globally optimal element is often intractable, many practical learning systems instead hillclimb to a loca ..."
Abstract

Cited by 12 (2 self)
 Add to MetaCart
Many learning systems search through a space of possible performance elements, seeking an element whose expected utility, over the distribution of problems, is high. As the task of finding the globally optimal element is often intractable, many practical learning systems instead hillclimb to a local optimum. Unfortunately, even this is problematic as the learner typically does not know the underlying distribution of problems, which it needs to determine an element's expected utility. This paper addresses the task of approximating this hillclimbing search when the utility function can only be estimated by sampling. We present a general algorithm, palo, that returns an element that is, with provably high probability, essentially a local optimum. We then demonstrate the generality of this algorithm by presenting three distinct applications, that respectively find an element whose efficiency, accuracy or completeness is nearly optimal. These results suggest approaches to solving the util...
StateAggregation Algorithms for Learning Probabilistic Models for Robot Control
, 2002
"... This thesis addresses the problem of learning probabilistic representations of dynamical systems with nonlinear dynamics and hidden state in the form of partially observable Markov decision process (POMDP) models, with the explicit purpose of using these models for robot control. In contrast to the ..."
Abstract

Cited by 7 (1 self)
 Add to MetaCart
This thesis addresses the problem of learning probabilistic representations of dynamical systems with nonlinear dynamics and hidden state in the form of partially observable Markov decision process (POMDP) models, with the explicit purpose of using these models for robot control. In contrast to the usual approach to learning probabilistic models, which is based on iterative adjustment of probabilities so as to improve the likelihood of the observed data, the algorithms proposed in this thesis take a different approach  they reduce the learning problem to that of state aggregation by clustering in an embedding space of delayed coordinates, and subsequently estimating transition probabilities between aggregated states (clusters). This approach has close ties to the dominant methods for system identification in the field of control engineering, although the characteristics of POMDP models require very different algorithmic solutions.
The Complexity of Revising Logic Programs
 The Journal of Logic Programming
, 1999
"... A rulebased program will return a set of answers to each query. An impure program, which includes the Prolog cut "!" and "not(\Delta)" operators, can return different answers if its rules are reordered. There are also many reasoning systems that return only the first answer found for each query; t ..."
Abstract

Cited by 4 (1 self)
 Add to MetaCart
A rulebased program will return a set of answers to each query. An impure program, which includes the Prolog cut "!" and "not(\Delta)" operators, can return different answers if its rules are reordered. There are also many reasoning systems that return only the first answer found for each query; these first answers, too, depend on the rule order, even in pure rulebased systems. A theory revision algorithm, seeking a revised rulebase whose expected accuracy, over the distribution of queries, is optimal, should therefore consider modifying the order of the rules. This paper first shows that a polynomial number of training "labeled queries" (each a query paired with its correct answer) provides the distribution information necessary to identify the optimal ordering. It then proves, however, that the task of determining which ordering is optimal, once given this distributional information, is intractable even in trivial situations; e.g., even if each query is an atomic literal, we are...
The Automated Refinement of a Requirements Domain Theory
 Journal of Automated Software Enginnering, Special Issue on Inductive Programming
, 1999
"... The specification and management of requirements is widely considered to be one of the most important yet most problematic phases in software engineering. In some applications, such as in safety critical areas or knowledgebased systems, the construction of a requirements domain theory is regarde ..."
Abstract

Cited by 3 (2 self)
 Add to MetaCart
The specification and management of requirements is widely considered to be one of the most important yet most problematic phases in software engineering. In some applications, such as in safety critical areas or knowledgebased systems, the construction of a requirements domain theory is regarded as an important part of this phase. Building and maintaining such a domain theory, however, requires a large investment and a range of powerful validation and maintenance tools. The area of `theory refinement' is concerned with the use of training data to automatically change an existing theory so that it better fits the data. Theory refinement techniques have not been extensively used in applications because of the problems in scaling up their underlying algorithms. In this paper we describe an environment for validating and maintaining a requirements domain theory written in a customised form of manysorted logic. The environment has been used for several years to maintain a theo...
A Case Study in the Use of Theory Revision in Requirements Validation
 In Machine Learning: Proceedings of the 15th International Conference Shavlik, J (Ed
, 1998
"... Research emanating from Artificial Intelligence has throughout its history contributed to techniques and ideas in Software Engineering. We describe in this paper a case study showing the use of theory revision to the refinement of a formally specified requirements model. In a previous project we wer ..."
Abstract

Cited by 3 (3 self)
 Add to MetaCart
Research emanating from Artificial Intelligence has throughout its history contributed to techniques and ideas in Software Engineering. We describe in this paper a case study showing the use of theory revision to the refinement of a formally specified requirements model. In a previous project we were contracted to create a precise model of the complex criteria governing the separation of aircraft profiles in Atlantic Airspace. During that work it became clear that the (automated) validation of the model was of the utmost importance, and in our current project we have used machine learning tools to provide extra support in bug identification, bug removal and maintenance of such a requirements model. In this paper we give an overview of the domain, identify a relevant learning bias which makes search for revisions tractable, and describe a systematic approach for the application of theory revision to such a model. We illustrate the approach with results of experiments where theory revisi...
New Horn revision algorithms
 Journal of Machine Learning Research
, 2005
"... A revision algorithm is a learning algorithm that identifies the target concept, starting from an initial concept. Such an algorithm is considered efficient if its complexity (in terms of the measured resource) is polynomial in the syntactic distance between the initial and the target concept, but o ..."
Abstract

Cited by 3 (0 self)
 Add to MetaCart
A revision algorithm is a learning algorithm that identifies the target concept, starting from an initial concept. Such an algorithm is considered efficient if its complexity (in terms of the measured resource) is polynomial in the syntactic distance between the initial and the target concept, but only polylogarithmic in the number of variables in the universe. We give efficient revision algorithms in the model of learning with equivalence and membership queries. The algorithms work in a general revision model where both deletion and addition revision operators are allowed. In this model one of the main open problems is the efficient revision of Horn formulas. Two revision algorithms are presented for special cases of this problem: for depth1 acyclic Horn formulas, and for definite Horn formulas with unique heads.
Grammar Approximation by Representative Sublanguage: A New Model for Language Learning
"... We propose a new language learning model that learns a syntacticsemantic grammar from a small number of natural language strings annotated with their semantics, along with basic assumptions about natural language syntax. We show that the search space for grammar induction is a complete grammar latt ..."
Abstract

Cited by 2 (1 self)
 Add to MetaCart
We propose a new language learning model that learns a syntacticsemantic grammar from a small number of natural language strings annotated with their semantics, along with basic assumptions about natural language syntax. We show that the search space for grammar induction is a complete grammar lattice, which guarantees the uniqueness of the learned grammar. 1
Theory revision with queries: Horn, readonce, and parity formulas
 Artificial Intelligence
, 2004
"... A theory, in this context, is a Boolean formula; it is used to classify instances, or truth assignments. Theories can model realworld phenomena, and can do so more or less correctly. The theory revision, or concept revision, problem is to correct a given, roughly correct concept. This problem is co ..."
Abstract

Cited by 2 (0 self)
 Add to MetaCart
A theory, in this context, is a Boolean formula; it is used to classify instances, or truth assignments. Theories can model realworld phenomena, and can do so more or less correctly. The theory revision, or concept revision, problem is to correct a given, roughly correct concept. This problem is considered here in the model of learning with equivalence and membership queries. A revision algorithm is considered efficient if the number of queries it makes is polynomial in the revision distance between the initial theory and the target theory, and polylogarithmic in the number of variables and the size of the initial theory. The revision distance is the minimal number of syntactic revision operations, such as the deletion or addition of literals, needed to obtain the target theory from the initial theory. Efficient
Towards the Automated Debugging and Maintenance of Logicbased Requirements Models
 In ASE '98: Proceedings of the 13th IEEE International Conference on Automated Software Engineering
, 1998
"... In this paper we describe a tools environment which automates the validation and maintenance of a requirements model written in manysorted first order logic. We focus on: a translator, that produces an executable form of the model; blame assignment functions, which input batches of misclassified t ..."
Abstract

Cited by 1 (1 self)
 Add to MetaCart
In this paper we describe a tools environment which automates the validation and maintenance of a requirements model written in manysorted first order logic. We focus on: a translator, that produces an executable form of the model; blame assignment functions, which input batches of misclassified tests (i.e. training examples) and output likely faulty parts of the model; and a theory reviser, which inputs the faulty parts and examples and outputs suggested revisions to the model. In particular, we concentrate on the problems encountered when applying these tools to a real application: a requirements model containing air traffic control separation standards, operating methods and airspace information. 1. Introduction A unifying theme in the research areas of knowledge engineering, requirements engineering and formal methods is the construction and validation of requirements models represented as formal systems (using languages such as RML [7]). Within the knowledge based system communi...
The Application of a Machine Learning Tool to the Validation of an Air Traffic Control Domain Theory
 In Proceedings of 12th IEEE International Conference on Tools with Artificial Intelligence (ICTAI
, 2000
"... In this paper we describe a project (IMPRESS) which utilised a Machine Learning Tool for the Validation of an Air Traffic Control Domain Theory. During the project, novel techniques were devised for the automated revision of general clause form theories using training examples. This technique involv ..."
Abstract

Cited by 1 (0 self)
 Add to MetaCart
In this paper we describe a project (IMPRESS) which utilised a Machine Learning Tool for the Validation of an Air Traffic Control Domain Theory. During the project, novel techniques were devised for the automated revision of general clause form theories using training examples. This technique involves focusing in on the parts of a theory which involve ordinal sorts, and applying geometrical revision operators to repair faulty component parts. The method is illustrated with experimental results obtained during the project.