Juan E. Vargas

A framework for the analysis of dynamic processes based on Bayesian networks and case-based reasoning

Abstract Bayesian networks are knowledge representation schemes that can capture probabilistic relationships among variables and perform probabilistic inference. Arrival of new evidence propagates through the network until all variables are updated. At the end of propagation, the network becomes a static snapshot representing the state of the domain for that particular time. This weakness in capturing temporal semantics has limited the use of Bayesian networks to domains in which time dependency is not a critical factor. This paper describes a framework that combines Bayesian networks and case-based reasoning to create a knowledge representation scheme capable of dealing with time-varying processes. Static Bayesian network topologies are learned from previously available raw data and from sets of constraints describing significant events. These constraints are defined as sets of variables assuming significant values. As new data are gathered, dynamic changes to the topology of a Bayesian network are assimilated using techniques that combine single-value decomposition and minimum distance length. The new topologies are capable of forecasting the occurrences of significant events given specific conditions and monitoring changes over time. Since environment problems are good examples of temporal variations, the problem of forecasting ozone levels in Mexico City was used to test this framework.

Expert system mixed-model assembly line scheduling

Abstract Consumer demand for variety forces manufacturers to produce models in families that can be adapted to include optional features. Scheduling assembly lines to produce these models is a complex task that requires the satisfaction of numerous constraints, such aa model selection and sequencing, lot sizes, models complexity, line interdependencies, lines capacity, and line changeover cost. This paper describes an expert system scheduler that uses a conflict resolution procedure driven by component changeover costs, intraline and interline constraints, and a set of heuristics that affect the production of the whole factory. Base models and their components are represented by object-oriented methods; rules accessing objects create production plans by resolving conflicts arising from the interactions between the numerous constraints. The resulting plans are then used to set up the production in the lines and to provide feedback to manufacturing and marketing.

A paradigm for building diagnostic expert systems by specializing generic device and reasoning models

This paper describes a generic schematic knowledge representation, acquisition, and manipulation system (SKRAM) applied to diagnostic problem solving in analog circuits. SKRAM is a general purpose declarative, schema-oriented system composed of five layers: (1) an object-oriented implementation layer, (2) a logical layer defining object features, i.e., relations and attributes, (3) an epistemological layer at which schemata are introduced as knowledge packages, (4) a semantic or conceptual layer where meaning is attached to schemata by defining an interpretation mechanism, and (5) a domain layer containing objects and schemata which represent knowledge of the application domain. SKRAM supports the acquisition and maintenance of general and specific knowledge and constructs models of the entities in the application domain, e.g., models of the structure and function of electronic circuits. An example of the use of the declarative representation in the domain of analog circuits is given and the architecture is contrasted with other architectures in similar and related domains.

Improving the scope of intelligent tutoring by adapting a case-based methodology through a distributed architecture

This paper describes an architecture for distributed case-based tutoring, called DICABTU, which provides an environment that facilitates cooperation among independent agents working together to provide highly individualized instruction. The fusion of these agents through a blackboard platform creates a distributed learning environment in which the most competent agents are called up to assist a student during a tutoring session. Following a curriculum derived from a node-based knowledge network, case-based reasoning is used to compose lessons at various levels of knowledge, to generate teaching materials, and to solve problems.

Similarity-based reasoning about diagnosis of analog circuits

This paper advocates the use of similarity-based reasoning to improve the efficacy of AI applications dealing with diagnosis of electronic circuits. A paradigm for similarity-based matching is presented, application of this paradigm to diagnosis of electronic circuits is discussed, and a system architecture that supports this concept is described.

Intelligent CAI in engineering: Knowledge representation strategies to facilitate model-based reasoning

This article describes an architecture for an intelligent computer‐assisted instruction (ICAI) system for use in engineering domains. A qualitative model is utilized for representing the physical characteristics and operating methodologies of the systems tutored; planning techniques are employed for instruction using a text‐based tutor; reasoning about the engineering model and about the student behavior during instruction is accomplished using a production system. Examples are drawn from two specific implementations: tutorials about (1) a hot water heater and (2) a power distribution system. Specific recommendations are given for implementing ICAI systems in engineering domains.

Home-based monitoring of cardiac patients

The paper describes an ongoing effort towards the design of systems for home based monitoring of cardiac patients. The goal is to develop systems that could support early discharge of cardiac patients and enable their home based recuperation. Such systems could act as a virtual extension of the ICU/CCU systems in the hospital, enabling physicians to monitor patients at home, and giving them the ability to discharge patients earlier while continuing to monitor their recuperation progress. This concept could result in direct health care cost savings and improvement in the quality of medical care, especially in rural areas with limited access to cardiologists. Our first step towards this goal is a system that can send cardiac data from the patient to the physician via the Internet. The physician can perform data analysis and send reports back to the patient over the Internet. In order to provide a truly home based monitoring system, we plan to add to the current system the ability to transmit real time cardiac and blood pressure events over cellular telephone networks.

Scale-guided object matching for case-based reasoning

Case-Based Reasoning (CBR) can be seen as a problem-solving paradigm that advocates the use of previous experiences to limit search spaces and to reduce opportunities for error repetition. In this paradigm, the case at hand is compared against former experiences to select from a set of possible courses of action the best one. A comparison method is required to ensure that the most resembling experience is, in fact, chosen to drive the problem-solving process. This paper discusses an object-oriented framework that provides a scale-guided measure of similarity between objects, and shows how this framework can be applied for case-based reasoning, drawing examples from device diagnosis.

DETECTING DEVICE FAILURES THROUGH SCALE-DRIVEN OBJECT MATCHING

Abstract This article presents a method for detecting device failures using scale-driven object matching and discusses the use of this method to facilitate object-oriented device diagnosis. A device-independent class called scale, introduced to hide the method from the objects, provides abstraction mechanisms that facilitate knowledge acquisition and system integration. Using this class, plans for diagnosis can be generated from the mismatches resulting from comparing prototypical devices with faulty devices.

Target Tracking with Bayesian Estimation

We present a Bayesian approach for multiple target tracking. Target location and velocity are deduced probabilistically through a sequence of continuous observations of amplitude and frequency made by Doppler Radar sensors.