We extend qualitative reasoning with estima tions of the relative likelihoods of the possible qu... more We extend qualitative reasoning with estima tions of the relative likelihoods of the possible qualitative behaviors. We estimate the likeli hoods by viewing the dynamics of a system as a Markov chain over its transition graph. This corresponds to adding probabilities to each of the transitions. The transition prob abilities follow directly from theoretical con siderations in simple cases. In the remaining cases, one must derive them empirically from numeric simulations, experiments, or subjec tive estimates. Once the transition probabili ties have been estimated, the standard theory of Markov chains provides extensive informa tion about asymptotic behavior, including a partition into persistent and transient states, the probabilities for ending up in each state, and settling times. Even rough estimates of transition probabilities provide useful qualita tive information about ultimate behaviors, as the analysis of many of these quantities is in sensitive to perturbations in the probabilities. The algorithms are straightforward and require time cubic in the number of qualitative states. The analysis also goes through for symbolic probability estimates, although at the price of exponential-time worst-case performance.
The Problem: Computer systems are under attack as they have never been before. Every day computer... more The Problem: Computer systems are under attack as they have never been before. Every day computer systems around the world are compromised: data is stolen, security is breached, services are denied. These attacks are enabled by flaws in our current operating system, application code and middleware. It would be nice to replace all this with new software that is based on provably impenetrable trusted computing base, but of course this isn't practical. We instead need a way to make systems behave in a more survivable fashion and to complete their critical tasks even when compromises have happened.
We extend qualitative reasoning with estimations of the relative likelihoods of the possible qual... more We extend qualitative reasoning with estimations of the relative likelihoods of the possible qualitative behaviors. We estimate the likelihoods by viewing the dynamics of a system as a Markov chain over its transition graph. This corresponds to adding probabilities to each of the transitions. The transition probabilities follow directly from theoretical considerations in simple cases. In the remaining cases, one must derive them empirically from numeric simulations, experiments, or subjective estimates. Once the transition probabilities have been estimated, the standard theory of Markov chains provides extensive information about asymptotic behavior, including a partition into persistent and transient states, the probabilities for ending up in each state, and settling times. Even rough estimates of transition probabilities provide useful qualitative information about ultimate behaviors, as the analysis of many of these quantities is insensitive to perturbations in the probabilities. The algorithms are straightforward and require time cubic in the number of qualitative states. The analysis also goes through for symbolic probability estimates, although at the price of exponential-time worst-case performance.
Abstract Affordability and responsiveness are two key requirements for the next generation of air... more Abstract Affordability and responsiveness are two key requirements for the next generation of aircraft support systems. Aircraft have to be repaired with minimum downtime and the support process has to be economical and efficient. The vision of autonomic logistics (AL) ...
Specification of objectives constitutes a central issue in knowledge representation for planning.... more Specification of objectives constitutes a central issue in knowledge representation for planning. Decision-theoretic approaches require that representations of objectives possess a firm semantics in terms of utility functions, yet provide the flexible compositionality needed for practical preference modeling for planning systems. Modularity, or separability in specification, is the key representational feature enabling this flexibility. In the context of utility specification, modularity corresponds exactly to well-known independence concepts from multiattribute utility theory, and leads directly to approaches for composing separate preference specifications. Ultimately, we seek to use this utilitytheoretic account to justify and improve existing mechanisms for specification of preference information, and to develop new representations exhibiting tractable specification and flexible composition of preference criteria. 1 REPRESENTING UTILITY FOR PLANNING As genera...
: Although many have proposed formal characterizations of belief structures as bases for rational... more : Although many have proposed formal characterizations of belief structures as bases for rational action, the problem of characterizing rational desires has attracted little attention. AI relies heavily on goal conditions interpreted (apparently) as absolute expressions of desirability, but these cannot express varying degrees of goal satisfaction or preferences among alternative goals. Our previous work provided a relative interpretation of goals as qualitative statements about preferability, all else equal. We extend that treatment to the comparison of arbitrary propositions, and develop a propositional logic of relative desire suitable for formalizing properties of planning and problem-solving methods. 1 Introduction Question your desires.---William Shakespeare Standard theories of rational action take decisions of the agent to depend on beliefs about the relative desirability of the results of its available actions [2, 3]. The predominant approach to planning in artificial intel...
Decision-theoretic preferences specify the relative desirability of all possible outcomes of alte... more Decision-theoretic preferences specify the relative desirability of all possible outcomes of alternative plans. In order to express general patterns of preference holding in a domain, we require a language that can refer directly to preferences over classes of outcomes as well as individuals. We present the basic concepts of a theory of meaning for such generic comparatives to facilitate their incremental capture and exploitation in automated reasoning systems. Our semantics lifts comparisons of individuals to comparisons of classes "other things being equal" by means of contextual equivalences, equivalence relations among individuals that vary with the context of application. We discuss implications of the theory for representing preference information. 1 Introduction Decision-theoretic treatments of preferences represent the objectives of a decision maker by an ordering over the possible outcomes of available plans. In taking a decision-theoretic approach to planning, we...
Efficiency dictates that plans for large-scale distributed activities be revised incrementally, w... more Efficiency dictates that plans for large-scale distributed activities be revised incrementally, with parts of plans being revised only if the expected utility of identifying and revising the subplans improve on the expected utility of using the original plan. The problems of identifying and reconsidering the subplans affected by changed circumstances or goals are closely related to the problems of revising beliefs as new or changed information is gained. But the current techniques of reason maintenance--- the standard method for belief revision---choose revisions arbitrarily and enforce global notions of consistency and groundedness which may mean reconsidering all beliefs or plan elements at each step. We outline revision methods that revise only those beliefs and plans worth revising, and that tolerate incoherence and ungroundedness when these are judged less detrimental than a costly revision effort. 1 Introduction Planning is necessary for the organization of lar...
Decision-theoretic preferences specify the relative desirability of all possible outcomes of alte... more Decision-theoretic preferences specify the relative desirability of all possible outcomes of alternative plans. In order to express general patterns of preference holding in a domain, we require a language that can refer directly to preferences over classes of outcomes as well as individuals. We present the basic concepts of a theory of meaning for such generic comparatives to facilitate their incremental capture and exploitation in automated reasoning systems. Our semantics lifts comparisons of individuals to comparisons of classes "other things being equal" by means of contextual equivalences, equivalence relations among individuals that vary with the context of application. We discuss implications of the theory for representing preference information. 1 Introduction Decision-theoretic treatments of preferences represent the objectives of a decision maker by an ordering over the possible outcomes of available plans. In taking a decision-theoretic approach to planning, we...
Although many have proposed formal characterizations of belief structures as bases for rational a... more Although many have proposed formal characterizations of belief structures as bases for rational action, the problem of characterizing rational desires has attracted little attention. AI relies heavily on goal conditions interpreted (apparently) as absolute expressions of desirability, but these cannot express varying degrees of goal satisfaction or preferences among alternative goals. Our previous work provided a relative interpretation of goals as qualitative statements about preferability, all else equal. We extend that treatment to the comparison of arbitrary propositions, and develop a propositional logic of relative desire suitable for formalizing properties of planning and problem-solving methods.
Goals, as typically conceived in AI planning, provide an insufficient basis for choice of action,... more Goals, as typically conceived in AI planning, provide an insufficient basis for choice of action, and hence are deficient as the sole expression of an agent's objectives. Decision-theoretic utilities offer a more adequate basis, yet lack many of the computational advantages of goals. We provide a preferential semantics for goals that grounds them in decision theory and preserves the validity of some, but not all, common goal operations performed in planning. This semantic account provides a criterion for verifying the design of goal-based planning strategies, thus providing a new framework for knowledge-level analysis of planning systems. Planning to achieve goals In the predominant AI planning paradigm, planners construct plans designed to produce states satisfying particular conditions called goals. Each goal represents a partition of possible states of the world into those satisfying and those not satisfying the goal. Though planners use goals to guide their reasoning, the cr...
Several recent problem-solving programs have indicated improved methods for controlling program a... more Several recent problem-solving programs have indicated improved methods for controlling program actions. Some of these methods operate by analyzing the time-independent antecedent-consequent dependency relationships between the components of knowledge about the problem for solution. This paper is a revised version of a thesis proposal which indicates how a general system of automatically maintained dependency relationships can be used to effect many forms of control on reasoning in an antecedent reasoning framework. 1 Introduction A major problem encountered in building automatic problem-solving systems is the necessity of providing controls on program reasoning and action. When faced with this problem, however, some researchers have divided reasoning into the categories of antecedent (datadirected) and consequent (goal-directed) reasoning. They then claim that the problem of control is just that of deciding on a proper balance between antecedent and consequent representations o...
: The Intrusion Detection (ID) community has developed numerous proposals for languages with whic... more : The Intrusion Detection (ID) community has developed numerous proposals for languages with which to describe signatures of attacks on computers and networks. By and large, these languages provide means for describing sequences of specific events indicative of attacks through their presence or absence in the history of some computational system. This note argues from examples that meeting the needs of information warfare requires significantly extending the expressive capabilities of attack description languages. This work is supported by DARPA under contract F30602-99-1-0509. 1 Introduction Intrusion Detection (ID) research efforts have explored two main methods of recognizing intrusions or attacks on computer systems and networks, namely signature recognition and anomaly detection. In signature detection, one examines the events occurring in the system under guard to determine if this history matches "signature" patterns that characterize types of attacks. In anomaly de...
Bayesian Neural Networks for Genetic Association Studies of Complex Disease
Discovering causal genetic variants from large genetic association studies poses many difficult c... more Discovering causal genetic variants from large genetic association studies poses many difficult challenges. Assessing which genetic markers are involved in determining trait status is a computationally demanding task, especially in the presence of gene-gene interactions. A non-parametric Bayesian approach in the form of a Bayesian neural network is proposed for use in analyzing genetic association studies. Demonstrations on synthetic and real data reveal they are able to efficiently and accurately determine which variants are involved in determining case-control status. Using graphics processing units (GPUs) the time needed to build these models is decreased by several orders of magnitude. In comparison with commonly used approaches for detecting genetic interactions, Bayesian neural networks perform very well across a broad spectrum of possible genetic relationships while having the computational efficiency needed to handle large datasets.
This report contains most of the text of a proposal we submitted in March, 1994 to ARPA in respon... more This report contains most of the text of a proposal we submitted in March, 1994 to ARPA in response to BAA 94-13, for the Health Information Infrastructure Program. This report differs from the proposal only in its organization and in omitting budgetary and administrative details. To a large extent, we have retained the language typical of proposals, and hope for the reader's patience with its tone
We extend qualitative reasoning with estima tions of the relative likelihoods of the possible qu... more We extend qualitative reasoning with estima tions of the relative likelihoods of the possible qualitative behaviors. We estimate the likeli hoods by viewing the dynamics of a system as a Markov chain over its transition graph. This corresponds to adding probabilities to each of the transitions. The transition prob abilities follow directly from theoretical con siderations in simple cases. In the remaining cases, one must derive them empirically from numeric simulations, experiments, or subjec tive estimates. Once the transition probabili ties have been estimated, the standard theory of Markov chains provides extensive informa tion about asymptotic behavior, including a partition into persistent and transient states, the probabilities for ending up in each state, and settling times. Even rough estimates of transition probabilities provide useful qualita tive information about ultimate behaviors, as the analysis of many of these quantities is in sensitive to perturbations in the probabilities. The algorithms are straightforward and require time cubic in the number of qualitative states. The analysis also goes through for symbolic probability estimates, although at the price of exponential-time worst-case performance.
The Problem: Computer systems are under attack as they have never been before. Every day computer... more The Problem: Computer systems are under attack as they have never been before. Every day computer systems around the world are compromised: data is stolen, security is breached, services are denied. These attacks are enabled by flaws in our current operating system, application code and middleware. It would be nice to replace all this with new software that is based on provably impenetrable trusted computing base, but of course this isn't practical. We instead need a way to make systems behave in a more survivable fashion and to complete their critical tasks even when compromises have happened.
We extend qualitative reasoning with estimations of the relative likelihoods of the possible qual... more We extend qualitative reasoning with estimations of the relative likelihoods of the possible qualitative behaviors. We estimate the likelihoods by viewing the dynamics of a system as a Markov chain over its transition graph. This corresponds to adding probabilities to each of the transitions. The transition probabilities follow directly from theoretical considerations in simple cases. In the remaining cases, one must derive them empirically from numeric simulations, experiments, or subjective estimates. Once the transition probabilities have been estimated, the standard theory of Markov chains provides extensive information about asymptotic behavior, including a partition into persistent and transient states, the probabilities for ending up in each state, and settling times. Even rough estimates of transition probabilities provide useful qualitative information about ultimate behaviors, as the analysis of many of these quantities is insensitive to perturbations in the probabilities. The algorithms are straightforward and require time cubic in the number of qualitative states. The analysis also goes through for symbolic probability estimates, although at the price of exponential-time worst-case performance.
Abstract Affordability and responsiveness are two key requirements for the next generation of air... more Abstract Affordability and responsiveness are two key requirements for the next generation of aircraft support systems. Aircraft have to be repaired with minimum downtime and the support process has to be economical and efficient. The vision of autonomic logistics (AL) ...
Specification of objectives constitutes a central issue in knowledge representation for planning.... more Specification of objectives constitutes a central issue in knowledge representation for planning. Decision-theoretic approaches require that representations of objectives possess a firm semantics in terms of utility functions, yet provide the flexible compositionality needed for practical preference modeling for planning systems. Modularity, or separability in specification, is the key representational feature enabling this flexibility. In the context of utility specification, modularity corresponds exactly to well-known independence concepts from multiattribute utility theory, and leads directly to approaches for composing separate preference specifications. Ultimately, we seek to use this utilitytheoretic account to justify and improve existing mechanisms for specification of preference information, and to develop new representations exhibiting tractable specification and flexible composition of preference criteria. 1 REPRESENTING UTILITY FOR PLANNING As genera...
: Although many have proposed formal characterizations of belief structures as bases for rational... more : Although many have proposed formal characterizations of belief structures as bases for rational action, the problem of characterizing rational desires has attracted little attention. AI relies heavily on goal conditions interpreted (apparently) as absolute expressions of desirability, but these cannot express varying degrees of goal satisfaction or preferences among alternative goals. Our previous work provided a relative interpretation of goals as qualitative statements about preferability, all else equal. We extend that treatment to the comparison of arbitrary propositions, and develop a propositional logic of relative desire suitable for formalizing properties of planning and problem-solving methods. 1 Introduction Question your desires.---William Shakespeare Standard theories of rational action take decisions of the agent to depend on beliefs about the relative desirability of the results of its available actions [2, 3]. The predominant approach to planning in artificial intel...
Decision-theoretic preferences specify the relative desirability of all possible outcomes of alte... more Decision-theoretic preferences specify the relative desirability of all possible outcomes of alternative plans. In order to express general patterns of preference holding in a domain, we require a language that can refer directly to preferences over classes of outcomes as well as individuals. We present the basic concepts of a theory of meaning for such generic comparatives to facilitate their incremental capture and exploitation in automated reasoning systems. Our semantics lifts comparisons of individuals to comparisons of classes "other things being equal" by means of contextual equivalences, equivalence relations among individuals that vary with the context of application. We discuss implications of the theory for representing preference information. 1 Introduction Decision-theoretic treatments of preferences represent the objectives of a decision maker by an ordering over the possible outcomes of available plans. In taking a decision-theoretic approach to planning, we...
Efficiency dictates that plans for large-scale distributed activities be revised incrementally, w... more Efficiency dictates that plans for large-scale distributed activities be revised incrementally, with parts of plans being revised only if the expected utility of identifying and revising the subplans improve on the expected utility of using the original plan. The problems of identifying and reconsidering the subplans affected by changed circumstances or goals are closely related to the problems of revising beliefs as new or changed information is gained. But the current techniques of reason maintenance--- the standard method for belief revision---choose revisions arbitrarily and enforce global notions of consistency and groundedness which may mean reconsidering all beliefs or plan elements at each step. We outline revision methods that revise only those beliefs and plans worth revising, and that tolerate incoherence and ungroundedness when these are judged less detrimental than a costly revision effort. 1 Introduction Planning is necessary for the organization of lar...
Decision-theoretic preferences specify the relative desirability of all possible outcomes of alte... more Decision-theoretic preferences specify the relative desirability of all possible outcomes of alternative plans. In order to express general patterns of preference holding in a domain, we require a language that can refer directly to preferences over classes of outcomes as well as individuals. We present the basic concepts of a theory of meaning for such generic comparatives to facilitate their incremental capture and exploitation in automated reasoning systems. Our semantics lifts comparisons of individuals to comparisons of classes "other things being equal" by means of contextual equivalences, equivalence relations among individuals that vary with the context of application. We discuss implications of the theory for representing preference information. 1 Introduction Decision-theoretic treatments of preferences represent the objectives of a decision maker by an ordering over the possible outcomes of available plans. In taking a decision-theoretic approach to planning, we...
Although many have proposed formal characterizations of belief structures as bases for rational a... more Although many have proposed formal characterizations of belief structures as bases for rational action, the problem of characterizing rational desires has attracted little attention. AI relies heavily on goal conditions interpreted (apparently) as absolute expressions of desirability, but these cannot express varying degrees of goal satisfaction or preferences among alternative goals. Our previous work provided a relative interpretation of goals as qualitative statements about preferability, all else equal. We extend that treatment to the comparison of arbitrary propositions, and develop a propositional logic of relative desire suitable for formalizing properties of planning and problem-solving methods.
Goals, as typically conceived in AI planning, provide an insufficient basis for choice of action,... more Goals, as typically conceived in AI planning, provide an insufficient basis for choice of action, and hence are deficient as the sole expression of an agent's objectives. Decision-theoretic utilities offer a more adequate basis, yet lack many of the computational advantages of goals. We provide a preferential semantics for goals that grounds them in decision theory and preserves the validity of some, but not all, common goal operations performed in planning. This semantic account provides a criterion for verifying the design of goal-based planning strategies, thus providing a new framework for knowledge-level analysis of planning systems. Planning to achieve goals In the predominant AI planning paradigm, planners construct plans designed to produce states satisfying particular conditions called goals. Each goal represents a partition of possible states of the world into those satisfying and those not satisfying the goal. Though planners use goals to guide their reasoning, the cr...
Several recent problem-solving programs have indicated improved methods for controlling program a... more Several recent problem-solving programs have indicated improved methods for controlling program actions. Some of these methods operate by analyzing the time-independent antecedent-consequent dependency relationships between the components of knowledge about the problem for solution. This paper is a revised version of a thesis proposal which indicates how a general system of automatically maintained dependency relationships can be used to effect many forms of control on reasoning in an antecedent reasoning framework. 1 Introduction A major problem encountered in building automatic problem-solving systems is the necessity of providing controls on program reasoning and action. When faced with this problem, however, some researchers have divided reasoning into the categories of antecedent (datadirected) and consequent (goal-directed) reasoning. They then claim that the problem of control is just that of deciding on a proper balance between antecedent and consequent representations o...
: The Intrusion Detection (ID) community has developed numerous proposals for languages with whic... more : The Intrusion Detection (ID) community has developed numerous proposals for languages with which to describe signatures of attacks on computers and networks. By and large, these languages provide means for describing sequences of specific events indicative of attacks through their presence or absence in the history of some computational system. This note argues from examples that meeting the needs of information warfare requires significantly extending the expressive capabilities of attack description languages. This work is supported by DARPA under contract F30602-99-1-0509. 1 Introduction Intrusion Detection (ID) research efforts have explored two main methods of recognizing intrusions or attacks on computer systems and networks, namely signature recognition and anomaly detection. In signature detection, one examines the events occurring in the system under guard to determine if this history matches "signature" patterns that characterize types of attacks. In anomaly de...
Bayesian Neural Networks for Genetic Association Studies of Complex Disease
Discovering causal genetic variants from large genetic association studies poses many difficult c... more Discovering causal genetic variants from large genetic association studies poses many difficult challenges. Assessing which genetic markers are involved in determining trait status is a computationally demanding task, especially in the presence of gene-gene interactions. A non-parametric Bayesian approach in the form of a Bayesian neural network is proposed for use in analyzing genetic association studies. Demonstrations on synthetic and real data reveal they are able to efficiently and accurately determine which variants are involved in determining case-control status. Using graphics processing units (GPUs) the time needed to build these models is decreased by several orders of magnitude. In comparison with commonly used approaches for detecting genetic interactions, Bayesian neural networks perform very well across a broad spectrum of possible genetic relationships while having the computational efficiency needed to handle large datasets.
This report contains most of the text of a proposal we submitted in March, 1994 to ARPA in respon... more This report contains most of the text of a proposal we submitted in March, 1994 to ARPA in response to BAA 94-13, for the Health Information Infrastructure Program. This report differs from the proposal only in its organization and in omitting budgetary and administrative details. To a large extent, we have retained the language typical of proposals, and hope for the reader's patience with its tone
Uploads
Papers by Jon Doyle