Models to predict naturalness and image quality for images containing three memory colors: sky, grass, and skin

When evaluating the image quality, people mostly would like to concentrate on the color appearance of memory objects, representing the naturalness and reality of the image scene. Generally, an image with objects which have perfect memory colors reproduction will give natural and harmonious feelings. Many previous studies had proved the critical role of naturalness in image quality assessment, but it was still tough to scale the image naturalness precisely. In this study, natural images with blue sky, green grass, and skin colors were selected and partially rendered to develop the model of preference and naturalness of typical memory colors. A psychophysical experiment was conducted to collect the visual data of these images. Afterward, the psychophysical data were used to build the preference models and naturalness models, respectively. The models were then compared with previous studies. Results showed that the new models could accurately predict the preference and naturalness of target memory colors.

Preferences Offset Stressors in Determining Survivability

Pure stress-based models of health predict that accumulated stresses promote illnesses that result in lower life expectancies. However, in mixed subjective preference models, effects of stress, seen as negative preference, can potentially be offset by achievement of positive personal preferences. According to mixed preference models, preventive medicine strategies for promoting health can operate either by alleviating stress or by enhancing subjective preferences. Over many decades we developed a theory and practice of rational, psychoacoustically-driven architectural acoustic design of concert halls.1-3 The theory incorporates both negative acoustic annoyance attributes (stressors, negative preferences) and positive ones. Using self-assessment surveys of 30 dialysis patients in Kobe, Japan, we used the methodology to assess the effects of subjective preferences on delaying onset of dialysis treatment (dialysis onset age, DOA). Hayashi’s multivariate regression method (I) for nonparametric data 5,8 was used to estimate effects of reported factors. Of these, six factors proved predictive of DOA (p-values): better or worse interpersonal relations (0.003), decades of full-time work (0.050), alcohol consumption (0.031) according to individual preference, present noisy home environment (0.090), other pollution (0.060), smoking (0.115). Other factors were either weakly- or un- correlated: hospitalizations, house moves, past hypertension, proteinuria, sex, pet ownership, presence of bad odors, past noise pollution. Preventative measures that enhance subjective preferences may thus delay the need for dialysis.

Development of personalized online systems for web search, recommendations, and e-commerce

Personalized online systems for Web search, news recommendation, and e-commerce are developed. The process of personalization of online systems consists of three main steps: determining a user's needs, classifying products or services, and matching the user's needs with suitable products or services. A multi-feature based method to automatically classify Web pages into categories of topics hierarchically representing the Web pages is proposed. An approach to modeling and quantifying a user's interests and preferences using the user's Web navigational data is presented. The approach is based on the premise that frequently visiting certain types of content or Web sites indicates that the user is interested in related content or retrieving information from those sites. A personalized search system utilizing a Web user's interest, preference and search context models is developed. A Web user's interest and preference models are constructed and updated by analyzing the user's navigational data and automatically classifying Web pages. A user's search context model is used to determine how the user's interest and preference models impact on his or her search behavior. An algorithm to re-rank search results generated by a conventional search engine is designed to provide a personalized Web search service. A hybrid recommender system of personalized recommendation of news on the Web is developed. Based on the similarities between Web pages and users' models of interest and preference, the Web pages are recommended to the users who are likely interested in the related topics. Moreover, the technique of collaborative filtering is employed, which aims to choose the trusted users and incorporate machine intelligence combined with human efforts. Once trusted users are determined, their behavior on the Web is considered as the manual recommendation part of the system. A method of classifying Web customers for planning customized e-marketing is proposed. The proposed e-marketing approach can be divided into four steps: determining a customer's general interest model, ascertaining a customer's local browsing model, classifying Web customers, and designing a personalized marketing and promotion plan for e-commerce based on the customer classification. Various experiments are carried out to demonstrate the effectiveness of the proposed approaches and systems.

Development of personalized online systems for web search, recommendations, and e-commerce

Personalized online systems for Web search, news recommendation, and e-commerce are developed. The process of personalization of online systems consists of three main steps: determining a user's needs, classifying products or services, and matching the user's needs with suitable products or services. A multi-feature based method to automatically classify Web pages into categories of topics hierarchically representing the Web pages is proposed. An approach to modeling and quantifying a user's interests and preferences using the user's Web navigational data is presented. The approach is based on the premise that frequently visiting certain types of content or Web sites indicates that the user is interested in related content or retrieving information from those sites. A personalized search system utilizing a Web user's interest, preference and search context models is developed. A Web user's interest and preference models are constructed and updated by analyzing the user's navigational data and automatically classifying Web pages. A user's search context model is used to determine how the user's interest and preference models impact on his or her search behavior. An algorithm to re-rank search results generated by a conventional search engine is designed to provide a personalized Web search service. A hybrid recommender system of personalized recommendation of news on the Web is developed. Based on the similarities between Web pages and users' models of interest and preference, the Web pages are recommended to the users who are likely interested in the related topics. Moreover, the technique of collaborative filtering is employed, which aims to choose the trusted users and incorporate machine intelligence combined with human efforts. Once trusted users are determined, their behavior on the Web is considered as the manual recommendation part of the system. A method of classifying Web customers for planning customized e-marketing is proposed. The proposed e-marketing approach can be divided into four steps: determining a customer's general interest model, ascertaining a customer's local browsing model, classifying Web customers, and designing a personalized marketing and promotion plan for e-commerce based on the customer classification. Various experiments are carried out to demonstrate the effectiveness of the proposed approaches and systems.

The Role of Model Selection in Preference Learning

Learning preferences of an agent requires choosing which preference representation to use. This formalism should be expressive enough to capture a significant part of the agent's preferences. Selecting the right formalism is generally not easy, as we have limited access to the way the agent makes her choices. It is then important to understand how ``universal" particular preference representation formalisms are, that is, whether they can perform well in learning preferences of agents with a broad spectrum of preference orders. In this paper, we consider several preference representation formalisms from this perspective: lexicographic preference models, preference formulas, sets of (ranked) preference formulas, and neural networks. We find that the latter two show a good potential as general preference representation formalisms. We show that this holds true when learning preferences of a single agent but also when learning models to represent consensus preferences of a group of agents.

Interdependent Altruistic Preference Models

Altruistic preferences or the desire to improve the well‐being of others even at one’s own expense can be difficult to incorporate into traditional value and utility models. It is straightforward to construct a multiattribute preference structure for one decision maker that includes the outcomes experienced by others. However, when multiple individuals incorporate one another’s well‐being into their decision making, this creates complex interdependencies that must be resolved before the preference models can be applied. We provide representation theorems for additive altruistic value functions for two-person, n-person, and group outcomes in which multiple individuals are altruistic. We find that in most cases it is possible to resolve the preference interdependencies and that modeling the preferences of altruistic individuals and groups is tractable.

Rank Aggregation via Heterogeneous Thurstone Preference Models

We propose the Heterogeneous Thurstone Model (HTM) for aggregating ranked data, which can take the accuracy levels of different users into account. By allowing different noise distributions, the proposed HTM model maintains the generality of Thurstone's original framework, and as such, also extends the Bradley-Terry-Luce (BTL) model for pairwise comparisons to heterogeneous populations of users. Under this framework, we also propose a rank aggregation algorithm based on alternating gradient descent to estimate the underlying item scores and accuracy levels of different users simultaneously from noisy pairwise comparisons. We theoretically prove that the proposed algorithm converges linearly up to a statistical error which matches that of the state-of-the-art method for the single-user BTL model. We evaluate the proposed HTM model and algorithm on both synthetic and real data, demonstrating that it outperforms existing methods.

A Preference-Based Approach to Assess a Component’s Design Readiness for Additive Manufacturing

Due to the benefits associated with additive manufacturing (AM), there are increasingly more opportunities to leverage AM to enable the fabrication of components that were previously made using conventional techniques such as subtractive manufacturing or casting. To support this transition, it is critical to be able to rigorously evaluate the technical and economic feasibility of additively manufacturing an existing component design. In order to support this evaluation, this paper presents a novel feasibility analysis that performs a multi-criteria assessment of AM readiness. Along with the development of these assessments, we also present a novel scoring approach for qualitatively and quantitatively evaluating the feasibility of each component assessment. This scoring approach, which leverages preference models from physical programing, introduces a flexible set of feasibility levels to assess the manufacturability capabilities of AM technologies. It also allows for the integration of a designer’s preferences toward the AM assessments, supporting the decision whether to utilize AM technologies or not. The presented feasibility analysis allows for both technical and economic benefits since it suggests only using AM for those products whose feasibility results are within suitable ranges. The details of the approach are illustrated using four sample parts with varying geometries. Experimental validation is also performed to demonstrate the robustness of the evaluation. Results obtained show the capability and generalizability of these approaches to analyze intricate geometries and provide useful decision support in AM feasibility analysis.