Vu A. Ha

Preference elicitation via theory refinement

We present an approach to elicitation of user preference models in which assumptions can be used to guide but not constrain the elicitation process. We demonstrate that when domain knowledge is available, even in the form of weak and somewhat inaccurate assumptions, significantly less data is required to build an accurate model of user preferences than when no domain knowledge is provided. This approach is based on the KBANN (Knowledge-Based Artificial Neural Network) algorithm pioneered by Shavlik and Towell (1989). We demonstrate this approach through two examples, one involves preferences under certainty, and the other involves preferences under uncertainty. In the case of certainty, we show how to encode assumptions concerning preferential independence and monotonicity in a KBANN network, which can be trained using a variety of preferential information including simple binary classification. In the case of uncertainty, we show how to construct a KBANN network that encodes certain types of dominance relations and attitude toward risk. The resulting network can be trained using answers to standard gamble questions and can be used as an approximate representation of a persons preferences. We empirically evaluate our claims by comparing the KBANN networks with simple backpropagation artificial neural networks in terms of learning rate and accuracy. For the case of uncertainty, the answers to standard gamble questions used in the experiment are taken from an actual medical data set first used by Miyamoto and Eraker (1988). In the case of certainty, we define a measure to which a set of preferences violate a domain theory, and examine the robustness of the KBANN network as this measure of domain theory violation varies.

Similarity of personal preferences: theoretical foundations and empirical analysis

We study the problem of defining similarity measures on preferences from a decision-theoretic point of view. We propose a similarity measure, called probabilistic distance, that originates from the Kendalls tau function, a well-known concept in the statistical literature. We compare this measure to other existing similarity measures on preferences. The key advantage of this measure is its extensibility to accommodate partial preferences and uncertainty. We develop efficient methods to compute this measure, exactly or approximately, under all circumstances. These methods make use of recent advances in the area of Markov chain Monte Carlo simulation. We discuss two applications of the probabilistic distance: in the construction of the Decision-Theoretic Video Advisor (DIVA), and in robustness analysis of a theory refinement technique for preference elicitation.

Geometric foundations for interval-based probabilities

The need to reason with imprecise probabilities arises in a wealth of situations ranging from pooling of knowledge from multiple experts to abstraction-based probabilistic planning. Researchers have typically represented imprecise probabilities using intervals and have developed a wide array of different techniques to suit their particular requirements. In this paper we provide an analysis of some of the central issues in representing and reasoning with interval probabilities. At the focus of our analysis is the probability cross-product operator and its interval generalization, the cc-operator. We perform an extensive study of these operators relative to manipulation of sets of probability distributions. This study provides insight into the sources of the strengths and weaknesses of various approaches to handling probability intervals. We demonstrate the application of our results to the problems of inference in interval Bayesian networks and projection and evaluation of abstract probabilistic plans.

Feature-based decomposition of inductive proofs applied to real-time avionics software: an experience report

The hardware and software in modern aircraft control systems are good candidates for verification using formal methods: they are complex, safety-critical, and challenge the capabilities of test-based verification strategies. We have previously reported on our use of model checking to verify the time partitioning property of the Deos/spl trade/ real-time operating system for embedded avionics. The size and complexity of this system have limited us to analyzing only one configuration at a time. To overcome this limit and generalize our analysis to arbitrary configurations we have turned to theorem proving. This paper describes our use of the PVS theorem prover to analyze the Deos scheduler. In addition to our inductive proof of the time partitioning invariant, we present a feature-based technique for modeling state-transition systems and formulating inductive invariants. This technique facilitates an incremental approach to theorem proving that scales well to models of increasing complexity, and has the potential to be applicable to a wide range of problems.

Modeling user preferences via theory refinement

We present an approach to elicitation of user preference models in which assumptions can be used to guide but not constrain the elicitation process. We show how to encode assumptions concerning preferential independence and monotonicity in a Knowledge-Based Artificial Neural Network. We quantify the degree to which user preferences violate a set of assumptions. We empirically compare the KBANN network with an unbiased ANN in terms of learning rate and accuracy for preferences consistent and inconsistent with the assumptions. We go on to demonstrate how the technique can be used to learn a fine-grained preference structure from simple binary classification data.

Statistical verification of two non-linear real-time UAV controllers

We present a versatile statistical verification methodology and we illustrate different uses of this methodology on two examples of nonlinear real-time UAV controllers. The first example applies our statistical methodology to the verification of a computation time property for a software implementation of a high-performance controller as a function of controller state variable values. The second example illustrates our statistical verification methodology applied to finding verifiably safe flight envelopes for a class of maneuvers, again as a function of controller state variable values. We compare our approach to verification with other statistical techniques used for estimating execution times and controller performance. We close with candidate topics for future work.

Balancing safety against performance: tradeoffs in Internet security

All Internet-accessible computing systems are currently faced with incessant threats ranging from simple script-kiddies to highly sophisticated criminal enterprises. In response to these threats, sites must perform extensive intrusion monitoring. This intrusion monitoring can have significant costs in terms of bandwidth, computing power, storage space, and licensing fees. Furthermore, when exploits are detected, the victims must take actions that can consume further resources and compromise their objectives (e.g., by reducing e-commerce server throughput). In this paper, we explore techniques for modeling the costs and benefits of various security monitoring and response actions. Given these models and stochastic expectations about the types of attacks that a site is likely to face, our CIRCADIA (cooperative intelligent real-time control architecture for dynamic information assurance) automatic security control system is able to make real-time tradeoffs between the level of safety and security that is enforced, and the level of system resources/performance that are applied to the main computational objectives (e.g., e-commerce transactions). We show how CIRCADIA is able to dynamically adjust its security activities to account for changing threat profiles and objectives. The result: a continually-optimized balance of security-maintaining activity that reduces risk while still allowing the system to meet its goals.