Papers by Bruce D'ambrosio
Lazy Propagation and Independence of Causal Influence
Symbolic and Quantitative Approaches to Reasoning and Uncertainty, 1999
ABSTRACT
The Florida AI Research Society Conference, 1999
The efficiency of algorithms for probabilistic inference in Bayesian networks can be improved by ... more The efficiency of algorithms for probabilistic inference in Bayesian networks can be improved by exploiting independence of causal influence. The factorized rep- resentation of independence of causal influence offers a factorized decomposition of certain independence of causal influence models. We describe how lazy prop- agation - a junction tree based inference algorithm - easily can be extended to take advantage
Probabilistic Relational Models of On-line User Behavior
Wepropose,the usefulness of probabilistic relational methods,for modeling ,user behavior at web ,... more Wepropose,the usefulness of probabilistic relational methods,for modeling ,user behavior at web ,sites. Web logs (aka "click streams"), server logs, and other data sources, taken as datasets for traditional machine learning algorithms, violate the iid assumption of most algorithms. Requests ("clicks") are not independent within a session, sessions for a visitor are not independent of one another, and page types,
Uncertainty in Artificial Intelligence, 1999
The noisy-or and its generalization noisy max have been utilized to reduce the com plexity of kno... more The noisy-or and its generalization noisy max have been utilized to reduce the com plexity of knowledge acquisition. In this pa per, we present a new representation of noisy max that allows for efficient inference in gen eral Bayesian networks. Empirical studies show that our method is capable of com puting queries in well-known large medical networks, QMR-DT and CPCS, for which no previous exact inference method has been shown to perform well.
Uncertainty in Artificial Intelligence, 1992
We report on an experimental investigation into opportunities for parallelism in belief net infer... more We report on an experimental investigation into opportunities for parallelism in belief net inference. Specifically, we report on a study performed of the available parallelism, on hypercube style machines, of a set of ran domly generated belief nets, using factoring (SPI) style inference algorithms. Our results indicate that substantial speedup is available, but that it is available only through paral lelization of individual conformal product op erations, and depends critically on finding an appropriate factoring. We find negligible op portunity for parallelism at the topological, or clustering tree, level.
Uncertainty in Artificial Intelligence, 1993
Given a belief network with evidence, the task of finding the l most probable ex planations (MPE)... more Given a belief network with evidence, the task of finding the l most probable ex planations (MPE) in the belief network is that of identifying and ordering the l most probable instantiations of the non-evidence nodes of the belief network. Although many approaches have been proposed for solving this problem, most work only for restricted topologies (i.e., singly connected belief net works). In this paper, we will present a new approach for finding l MPEs in an arbitrary belief network. First, we will present an al gorithm for finding the MPE in a belief net work. Then, we will present a linear time al gorithm for finding the next MPE after find ing the first MPE. And finally, we will discuss the problem of fi nding the MPE for a subset of variables of a belief network, and show that the problem can be efficiently solved by this approach.
Security Situation Assessment and Response Evaluation (SSARE)
Proceedings DARPA Information Survivability Conference and Exposition II. DISCEX'01, 2000
A response to cyber attack is a decision made in the face of risk and uncertainty. Uncertainty, b... more A response to cyber attack is a decision made in the face of risk and uncertainty. Uncertainty, both in our understanding of the current situation and our capacity to predict exactly the results of alternate responses, requires the ability to entertain multiple hypotheses about the actual state of system security, attacker intent, and response effects. Risk management for catastrophic or near-catastrophic breaches of security or loss of service (either through system compromise or overly aggressive response) requires the evaluation of tradeoffs among competing objectives. Security Situation Assessment and Response Evaluation (SSARE) is a mixed-initiative computer software system for wide-area cyber attack detection, situation assessment, and response evaluation. SSARE is designed to detect a large-scale attack in progress, display an assessment of the situation, and identify effective responses, including automated context and risk-sensitive policy adaptations. The core of our technical approach is (I) development of attack, attacker, mission, systems, and infrastructure element models; (2) application of IET-developed information fusion and dynamic situation assessment technology, and (3) decision-theoretic evaluation of responses
Lazy Propagation and Independence of Causal Influence
Lecture Notes in Computer Science, 1999
ABSTRACT
The Symbolic Probabilistic Inference (SPI) Algorithm [D'Ambrosio, 19891 provides an efficient fra... more The Symbolic Probabilistic Inference (SPI) Algorithm [D'Ambrosio, 19891 provides an efficient framework for resolving general queries on a belief network. It applies the concept of dependency-directed backward search to probabilistic inference, and is incremental with respect to both queries and observations.
This final report summarizes technical achievements made for Phase I STTR project, entitled Auton... more This final report summarizes technical achievements made for Phase I STTR project, entitled Autonomous Distributed System - Multiple-Level Distributed Data Fusion for Future DADS Using Bayesian Network Technology -, during the period from 1 July 2001 to 15 January 2002. The future DADS, also known as mini-DADS or micro-DADS, is the next-generation DADS envisioned for 2020 littoral undersea threats. The most prominent feature of the future DADS is the elimination of a fiber line for an acoustic arrays for each of the current DADS sensor node, which poses a wide variety of technical challenges to implement distributed data fusion, including target detection, tracking, and classification. The Phase I efforts are to show technical feasibility of application of the distributed Bayesian network inference algorithms developed by IET and OSU in the past decade. Communication between nodes is conducted using Bayesian networks fragments to represent local information, and Bandwidth Agile Situ...
In this paper, we present the results of initial explorations into the application of relational ... more In this paper, we present the results of initial explorations into the application of relational model discovery methods to building comprehensive ecosystem models from data. Working with collaborators at the USGS Biological Resources Discipline and at the Environmental Protection Agency, we are engaged in two projects that apply relational probabilistic model discovery to building "community-level" models of ecosystems. A community-level ecosystem model is an integrated model of the ecosystem as a whole. The goal of our modeling effort is to aid domain scientists in gaining insight into data. Our preliminary work leads us to believe the method has tremendous promise. At the same time, we have encountered some limitations in existing methods. We briefly describe two projects and make some observations, particularly with respect to the development of synthetic, or derived, variables. We describe specific extensions we made to solve problems we encountered, and suggest eleme...
<title>Probabilistic tracking and real-time assessment of physical systems</title>
Knowledge-Based Artificial Intelligence Systems in Aerospace and Industry, 1994
ABSTRACT In order for long-range autonomous robots to be successful, they must be able to maintai... more ABSTRACT In order for long-range autonomous robots to be successful, they must be able to maintain accurate information about their location, available resources, and the state of critical components. We propose here a methodology that incorporates traditional, sensor-based tracking methods with discrete probabilistic representations of system state. Further, we extend the use of the Gaussian distribution to include a richer set of mathematical descriptions of system performance under specific failure conditions. The extended representations are then used to statistically test for these failure conditions by predicting the most likely values for observable parameters given the system state. This technique is then combined with first-order extended Kalman filtering to yield a probabilistic framework for tracking and fault detection in domains with nonlinear dynamics.
Proceedings of 9th International Parallel Processing Symposium, 1995
The computation complpxzty of the total probabiliiy mass of a leaf nodt of a geneinl Bayes networ... more The computation complpxzty of the total probabiliiy mass of a leaf nodt of a geneinl Bayes network can be exponential in t h e number of ancestor nodes of that leaf Il is a iuuell-known result that f o r a large cIiiss of networks, a number of m i n t e m s only linear in the number of ancestor nodes contributes about 67% of the total probabilrty mass. The problem of U!iyes net search is to yenprate only fhese high-mass mmterms. Wf inlrodui e U concurrpiit algarathm lor attempling this, based on converlinq the net into a concurrent process network Each parent node sends messages rontamtng p u rtial manternis lo child nodes. The novel idea is to prrorztzte these messages to give hzgher wezght to partial tvrms that are likely candadotes for znclvsioii in Il'e have implemented 1 L S S zts attributes.
Proceedings of the Seventh Conference on Uncertainty in Artificial Intelligence (1991)
This is the Proceedings of the Seventh Conference on Uncertainty in Artificial Intelligence, whic... more This is the Proceedings of the Seventh Conference on Uncertainty in Artificial Intelligence, which was held in Los Angeles, CA, July 13-15, 1991
The noisy-or and its generalization noisy-max have been utilized to reduce the complexity of know... more The noisy-or and its generalization noisy-max have been utilized to reduce the complexity of knowledge acquisition. In this paper, we present a new representation of noisy-max that allows for efficient inference in general Bayesian networks. Empirical studies show that our method is capable of computing queries in well-known large medical networks, QMR-DT and CPCS, for which no previous exact inference method has been shown to perform well.
IEEE Transactions on Systems, Man, and Cybernetics, 1995
Abstruct-Given a belief network with evidence, the task of finding the 1 most probable explanatio... more Abstruct-Given a belief network with evidence, the task of finding the 1 most probable explanations (MPE) in the belief network is that of identifying and ordering the 1 most probable instantiations of the non-evidence nodes. Although many approaches have been proposed for solving this problem, most work only for restricted topologies (e.& singly connected belief networks). In this paper we will present a framework, optimal factoring, for finding the 1 MPEs in arbitrary belief networks. Under this framework, efficiently finding the MPE in a belief network can be considered as the problem of finding an ordering of the distributions of the belief network and efficiently combining them. We will discuss the essence of the problem of finding the MPE, and present an optimal algorithm for singly connected belief networks and an efficient algorithm for multiply connected belief networks. We will also discuss the problem of finding the MPE for a subset of variables of a belief network under this framework.
UAI '91: Proceedings of the Seventh Annual Conference on Uncertainty in Artificial Intelligence, July 13-15, 1991, University of California at Los Angeles, Los Angeles, CA, USA
Uncertainty in Artificial Intelligence, 1991
Proceedings of the Eighth Conference on Uncertainty in Artificial Intelligence (1992)
ABSTRACT This is the Proceedings of the Eighth Conference on Uncertainty in Artificial Intelligen... more ABSTRACT This is the Proceedings of the Eighth Conference on Uncertainty in Artificial Intelligence, which was held in Stanford, CA, July 17-19, 1992
A partitioned ATMS
[1991] Proceedings. The Seventh IEEE Conference on Artificial Intelligence Application, 1991
ABSTRACT Describes an extension to deKleer&#39;s (1986) assumption-based truth maintenance sy... more ABSTRACT Describes an extension to deKleer&#39;s (1986) assumption-based truth maintenance system (ATMS) that shows promise in making large-scale problem solving more tractable. The approach assumes a problem can be decomposed into a hierarchically related set of subproblems, each of which is represented within its own ATMS partition. It is shown how this decomposition can be used to circumvent the exponential growth in the size of node labels which often occurs in attempting to apply the ATMS. The authors present some experimental results showing efficiency gains, and discuss the impact of partitioning on label consistency, soundness, minimality, and completeness
National Conference on Artificial Intelligence, 1990
The Symbolic Probabilistic Inference (SPI) Algorithm (D'Ambrosio, 19891 provides an efficient... more The Symbolic Probabilistic Inference (SPI) Algorithm (D'Ambrosio, 19891 provides an efficient framework for resolving general queries on a belief network. It applies the concept of dependency-directed backward search to probabilistic inference, and is incremental with respect to both queries and observations. Unlike most belief network algorithms, SPI is goal directed, performing only those calculations that are required to respond to
Uploads
Papers by Bruce D'ambrosio