Edward M. Hofstetter. / "December 9, 1959." Issued also as a thesis, M.I.T. Dept. of Electrical Engineering, August 24, 1959. / Bibliography: p. 35. / Army Signal Corps Contract DA36-039-sc-78108. Dept. of the Army Task 3-99-20-001 and Project 3-99-00-000.

We present a tree-structured architecture for supervised learning. The statistical model underlying the architecture is a hierarchical mixture model in which both the mixture coefficients and the mixture components are generalized linear models (GLIM's). Learning is treated as a maximum likelihood problem; in particular, we present an Expectation-Maximization (EM) algorithm for adjusting the parameters of the architecture. We also develop an on-line learning algorithm in which the parameters are updated incrementally. Comparative simulation results are presented in the robot dynamics domain.

Thesis (Ph. D.)--Ohio State University, 1983. / Includes vita. Includes bibliographical references (leaves 238-251). Available online via OhioLINK's ETD Center.

Thesis (Ph. D.)--Ohio State University, 2006. / Title from first page of PDF file. Includes bibliographical references (p. 132-152).

Thesis (Ph.D.)--University of Delaware, 2006. / Principal faculty advisors: Sandra M. Carberry and Keith S. Decker, Dept. of Computer & Information Sciences. Includes bibliographical references.

This thesis examines consumer behavior towards food risks in three different papers, focusing on two food concerns: genetically modified (GM) food and bovine spongiform encephalopathy (BSE). The first paper investigates the roles of different measures of trust on consumers stated choices for functional GM/nonGM canola oil products. These analyses show that consumers choices for GM/nonGM canola oil are influenced by both generalized trust and trust in food institutions. In general, trusting people are less likely to be in the group of respondents that can be characterized as being anti-GM; trusting people also tend to place a lower discount on the presence of a GM attribute. The second paper focuses on the modeling of consumers choices of foods with potential health and risk attributes. The analysis extends the linear compensatory utility model by allowing for use of attribute cutoffs in decision making. We find evidence that attribute cutoffs are commonly used by decision makers. Further, incorporating attribute cutoffs into the modeling of consumers choices significantly improved the model fit. This paper also examines a potential problem of endogeneity that may be associated with respondents self-reported cutoffs. Model estimates based on self-reported cutoffs differ substantially from those based on predicted cutoffs (where these are based on respondents demographic characteristics); potential reasons include the possibility that self-reported cutoffs may be endogenous. The third paper reports the impacts of habit and trust on consumers responses to a series of three BSE incidents in Canada. We observe that households reactions to the first two BSE events followed a similar pattern: households reduced their beef expenditure shares following the BSE announcements, but these subsequently recovered. We find that habit persistence reduced some households initial negative reactions to the first BSE incident, but that these households modified their beef consumption habits following recurring BSE incidents. Assessing the impacts of trust on households reactions to these BSE incidents, we find that trust tended to offset the negative effects of recurring BSE cases. / Agricultural and Resource Economics

Thesis (Ph.D.)--Kent State University, 2005. / Title from PDF t.p. (viewed Feb. 23, 2005). Advisor: Acar, William. Available online via OhioLINK's ETD Center. Keywords: strategy making; causal mapping; multi-agent systems; knowledge management; dialectical inquiry Includes bibliographical references (p. 111-116).

The purpose of this thesis is to integrate behavioral finance with market microstructure and financial decision-making. Specifically, I focus on two issues concerning the integration of psycho-physiological mechanisms and the informational content of prices in financial markets: firstly, the role of emotions in financial decision making and how as an adaptive mechanism, they show to be more suitable for survival than pure rationality (in an economic sense); and secondly, the empirical and theoretical testing of how cognitive illusions and polysemy affect the informational content of prices.

Are decision-makers, and public ones in particular, considering the spatial characteristics of territories when they are making decisions? Which elements should be included in computer-based systems supporting such decisions? These two questions summarise the issues geographic information system scientists are addressing. They also constitute the starting point of this investigation on the theme of errors and uncertainty in spatial decision support systems. While decision theories, methodologies and techniques abound, none are peculiar to spatial issues. At best, they are using spatial data and in a few cases they are applied to helping solve spatial problems. It has been recognised that spatial data need distinct management systems. Should it be the same for spatial decisions? By postulating that spatial decision is a specific field requiring its own treatise, this author initiates an original reflexion. Following a holistic approach, this theoretical work develops a conceptual decision model: the 'decisional fountain'. This model gives a coherent and integrated view on the various fields of decision-making. Based on the postulate and on the model developed, several theories and techniques are revisited in a truly spatial perspective. The ontology of spatial decision has been enriched by a variety of experiences. Dealing with drought at the European level, allocating water resources from the field scale to the country scale, regional spatial planning, performing strategic environmental assessment, conceptualising the parking of a vehicle, developing the spatial abilities of children, all have fed this conceptual research. Some of them are illustrating this manuscript. Taking a spatial information science perspective, this manuscript leads you towards a 'spatial decision' theory.

The amount of information involved in many decision making situations has increased dramatically in recent years and support of some kind is often needed. Consequently, fields like Business Intelligence (BI) and Decision Support Systems (DSS) have advanced. Decision analysis applications belong to the latter category and aim to support decision making activities in businesses and organizations, and provide more clearly structured decision material to use as a basis for decisions. In spite of a belief in their potential, their employment is still limited in practice, which could partly be attributed to the fact that they are incomplete to support decision processes sufficiently in real settings. At present, e.g., the specification and execution of the elicitation of input data is often left to the discretion of the user. Yet, this involves quite a few problematic elements and is of importance for the quality of the process as a whole. This thesis focuses on more practically useful elicitation of information in decision analysis applications than what is offered today. A process model emphasizing the importance of structured elicitation of adequate input data throughout decision processes is also suggested. In order to further define the problematic aspects of elicitation, three empirical studies were conducted. The problems with eliciting precise decision data suggests that using imprecise values within elicitation is a more realistic and useful approach to strive for. Based on theory and the findings of the studies, a weight elicitation method for imprecise statements and noisy input was formalized into the Cardinal Rank Ordering of Criteria (CROC) method. This method is both compatible with an adapted prescriptive decision making model, focused on a more structured elicitation component, as well as algorithms for dealing with such data. The CROC method was employed and validated in two real-life cases, which is not so common within decision analysis research. / Mängden information i många beslutssituationer har ökat markant under senare år och det finns ofta behov av någon form av stöd. Följaktligen har områden som Business Intelligence (BI) och Beslutsstödssystem (BSS) avancerat. Beslutsanalysverktyg tillhör den senare kategorin och syftar till att fungera som stöd vid beslutsfattande inom företag och organisationer och tillhandahålla mer strukturerat underlag för beslut. Trots en tro på deras potential, så är deras användande begränsat i praktiken, vilket delvis kan tillskrivas det faktum att de är inkompletta för att stödja beslutsprocesser i tillräcklig utsträckning i verkligheten. För närvarande förutsätts, t.ex. ofta att användaren själv klarar av att specificera och utföra utvinningen (eliciteringen) av input data. Detta involverar dock ett antal problematiska delar och dess kvalité är av vikt för hela processen. Denna avhandling fokuserar på mer praktiskt användbar elicitering av information i beslutsanalys-applikationer än vad som finns att tillgå idag. En processmodell som betonar vikten av strukturerad elicitering av adekvata indata genom hela beslutsprocessen föreslås också. För att ytterligare definiera de problematiska aspekterna av elicitering utfördes tre empiriska studier. Problemen med att utvinna precisa beslutsdata antyder att användandet av oprecisa värden inom elicitering är en mer realistisk och användbar ansats att sträva efter. Baserat på teori och resultaten av studierna formaliserades en vikteliciterings-metod för oprecisa utlåtanden och osäkra indata i Cardinal Rank Ordering of Criteria (CROC) metoden. Metoden är både kompatibel med en anpassad preskriptiv beslutsmodell fokuserad på en mer strukturerad eliciteringskomponent samt algoritmer för att hantera denna typ av data. CROC-metoden användes och validerades i två riktiga fall, vilket inte är så vanligt inom beslutsanalys forskning. / <p>At the time of the doctoral defense, the following papers were unpublished and had a status as follows: Paper 4: Accepted. Paper 7: Submitted. </p>