Bayesian Active Learning for Classification and Preference Learning

Information theoretic active learning has been widely studied for probabilistic models. For simple regression an optimal myopic policy is easily tractable. However, for other tasks and with more complex models, such as classification with nonparametric models, the optimal solution is harder to compute. Current approaches make approximations to achieve tractability. We propose an approach that expresses information gain in terms of predictive entropies, and apply this method to the Gaussian Process Classifier (GPC). Our approach makes minimal approximations to the full information theoretic objective. Our experimental performance compares favourably to many popular active learning algorithms, and has equal or lower computational complexity. We compare well to decision theoretic approaches also, which are privy to more information and require much more computational time. Secondly, by developing further a reformulation of binary preference learning to a classification problem, we extend our algorithm to Gaussian Process preference learning.

Predictive neural coding of reward preference involves dissociable responses in human ventral midbrain and ventral striatum.

Food preferences are acquired through experience and can exert strong influence on choice behavior. In order to choose which food to consume, it is necessary to maintain a predictive representation of the subjective value of the associated food stimulus. Here, we explore the neural mechanisms by which such predictive representations are learned through classical conditioning. Human subjects were scanned using fMRI while learning associations between arbitrary visual stimuli and subsequent delivery of one of five different food flavors. Using a temporal difference algorithm to model learning, we found predictive responses in the ventral midbrain and a part of ventral striatum (ventral putamen) that were related directly to subjects' actual behavioral preferences. These brain structures demonstrated divergent response profiles, with the ventral midbrain showing a linear response profile with preference, and the ventral striatum a bivalent response. These results provide insight into the neural mechanisms underlying human preference behavior.

Effect of several feeding stimulants on diet preference by juvenile gibel carp (Carassius auratus gibelio), fed diets with or without partial replacement of fish meal by meat and bone meal

The objectives of this work were to study the effects of several feeding stimulants on gibel carp fed diets with or without replacement of fish meal by meat and bone meal (MBM). The feeding stimulants tested were betaine, glycine, L-lysine, L-methionine, L-phenylalanine, and a commercial squid extract. Three inclusion levels were tested for each stimulant (0.18, 0.5%, and 1% for betaine and 0.1, 0.25 and 0.5% for the other stimulants). Two basal diets (40% crude protein) were used. one with 26% fish meal (FM), and the other with 21% fish meal and 6% MBM, Betaine at 0.1% in the fish meal group and at 0.5% in the meat and bone meal group was used in all experiments for comparison among stimulants. In the experiment on each stimulant, six tanks of fish were equally divided into two groups, one fed the FM diet, and the other fed the MBM diet. After 7 days' adaptation to the basal diet, in which the fish were fed to satiation twice a day, the fish were fed for another 7 days an equal mixture of diets containing varying levels of stimulants. Each diet contained a unique rare earth oxide as inert marker (Y2O3, Yb2O3, La2O3, Sm2O3 or Nd2O3). During the last 3 days of the experiment, faeces from each tank were collected. Preference for each diet was estimated based on the relative concentration of each marker in the faeces. Gibel carp fed the FM diet had higher intake than those fed the MBM diet, but the difference was significant only in the experiments on betaine, glycine and L-methionine. None of the feeding stimulants tested showed feeding enhancing effects in FM diets. All feeding stimulants showed feeding enhancing effects in MBM diets. and the optimum inclusion level was 0.5% for betaine, 0.1% for glycine, 0.25% for L-lysine, 0.1% for L-methionine. 0.25% For L-phenylalanine. and 0.1% for squid extract. The squid extract had the strongest stimulating effect among all the stimulants tested. (C) 2001 Elsevier Science B.V. All rights reserved.

Theoretical investigation on the phase stability of Nd2Fe14B and site preference of V, Cr, Mn, Zr and Nb

The stability of the excellent permanent magnetic compound Nd2Fe14B and substitution of Fe in the compound by V, Cr, Mn, Zr and Nb are investigated by using interatomic pair potentials which are converted from lattice-inversion method. Calculation shows that the substitution always makes the cell volume larger, and the increase of the volume is almost linear with substituent concentration. The calculated cohesive energy shows that the preferential order of substitution of Fe is Nb, V, Cr, Mn, Zr. Nevertheless, all the five substituting elements should most preferentially replace Fe in the j(2)' site, which has the greatest space among all six Fe sites. (C) 2005 Elsevier B.V. All rights reserved.

Natural Language Syntax and First Order Preference

We have argued elsewhere that first order inference can be made more efficient by using non-standard syntax for first order logic. In this paper we show how a fragment of English syntax under Montague semantics provides the foundation of a new inference procedure. This procedure seems more effective than corresponding procedures based on either classical syntax of our previously proposed taxonomic syntax. This observation may provide a functional explanation for some of the syntactic structure of English.

Finding optimal alternatives based on efficient comparative preference inference

Choosing the right or the best option is often a demanding and challenging task for the user (e.g., a customer in an online retailer) when there are many available alternatives. In fact, the user rarely knows which offering will provide the highest value. To reduce the complexity of the choice process, automated recommender systems generate personalized recommendations. These recommendations take into account the preferences collected from the user in an explicit (e.g., letting users express their opinion about items) or implicit (e.g., studying some behavioral features) way. Such systems are widespread; research indicates that they increase the customers' satisfaction and lead to higher sales. Preference handling is one of the core issues in the design of every recommender system. This kind of system often aims at guiding users in a personalized way to interesting or useful options in a large space of possible options. Therefore, it is important for them to catch and model the user's preferences as accurately as possible. In this thesis, we develop a comparative preference-based user model to represent the user's preferences in conversational recommender systems. This type of user model allows the recommender system to capture several preference nuances from the user's feedback. We show that, when applied to conversational recommender systems, the comparative preference-based model is able to guide the user towards the best option while the system is interacting with her. We empirically test and validate the suitability and the practical computational aspects of the comparative preference-based user model and the related preference relations by comparing them to a sum of weights-based user model and the related preference relations. Product configuration, scheduling a meeting and the construction of autonomous agents are among several artificial intelligence tasks that involve a process of constrained optimization, that is, optimization of behavior or options subject to given constraints with regards to a set of preferences. When solving a constrained optimization problem, pruning techniques, such as the branch and bound technique, point at directing the search towards the best assignments, thus allowing the bounding functions to prune more branches in the search tree. Several constrained optimization problems may exhibit dominance relations. These dominance relations can be particularly useful in constrained optimization problems as they can instigate new ways (rules) of pruning non optimal solutions. Such pruning methods can achieve dramatic reductions in the search space while looking for optimal solutions. A number of constrained optimization problems can model the user's preferences using the comparative preferences. In this thesis, we develop a set of pruning rules used in the branch and bound technique to efficiently solve this kind of optimization problem. More specifically, we show how to generate newly defined pruning rules from a dominance algorithm that refers to a set of comparative preferences. These rules include pruning approaches (and combinations of them) which can drastically prune the search space. They mainly reduce the number of (expensive) pairwise comparisons performed during the search while guiding constrained optimization algorithms to find optimal solutions. Our experimental results show that the pruning rules that we have developed and their different combinations have varying impact on the performance of the branch and bound technique.

CloudNeg: An autonomous multi issue negotiation system, with preference elicitation component, for trading cloud services

Cloud services provide its users with flexible resource provisioning. But in the current market, a user has to choose from a limited set of configurations at a fixed price. This paper presents an autonomous negotiation system termed CloudNeg for negotiating cloud services. CloudNeg provides buyers and sellers of cloud services with autonomous agents to negotiate on the specifications of a cloud instance, including price, on their behalf. These agents elicit their buyers’ time preferences and use them in negotiations. Further, this paper presents two artifacts: a negotiation algorithm and a prototype which together form CloudNeg.

Valuing maternity care: a comparison of stated preference methods with an application to cost-benefit analysis

This research investigates whether a reconfiguration of maternity services, which collocates consultant- and midwifery-led care, reflects demand and value for money in Ireland. Qualitative and quantitative research is undertaken to investigate demand and an economic evaluation is performed to evaluate the costs and benefits of the different models of care. Qualitative research is undertaken to identify women’s motivations when choosing place of delivery. These data are further used to inform two stated preference techniques: a discrete choice experiment (DCE) and contingent valuation method (CVM). These are employed to identify women’s strengths of preferences for different features of care (DCE) and estimate women’s willingness to pay for maternity care (CVM), which is used to inform a cost-benefit analysis (CBA) on consultant- and midwifery-led care. The qualitative research suggests women do not have a clear preference for consultant or midwifery-led care, but rather a hybrid model of care which closely resembles the Domiciliary Care In and Out of Hospital (DOMINO) scheme. Women’s primary concern during care is safety, meaning women would only utilise midwifery-led care when co-located with consultant-led care. The DCE also finds women’s preferred package of care closely mirrors the DOMINO scheme with 39% of women expected to utilise this service. Consultant- and midwifery-led care would then be utilised by 34% and 27% of women, respectively. The CVM supports this hierarchy of preferences where consultant-led care is consistently valued more than midwifery-led care – women are willing to pay €956.03 for consultant-led care and €808.33 for midwifery-led care. A package of care for a woman availing of consultant- and midwifery-led care is estimated to cost €1,102.72 and €682.49, respectively. The CBA suggests both models of care are cost-beneficial and should be pursued in Ireland. This reconfiguration of maternity services would maximise women’s utility, while fulfilling important objectives of key government policy.

Strong preference for decapod prey by the western rock lobster Panulirus cygnus

Western rock lobsters, Panulirus cygnus are an abundant benthic consumer distributed along the temperate west coast of Australia and constitute the largest single species fishery in Australia. As a dominant consumer, it is important to understand their predator-prey interactions as they can potentially exert strong trophic effects, and may influence ecosystem function as seen in other spiny lobster species. While previous field studies have focused on the diet composition of P. cygnus, this study investigated their preference for various benthic invertebrate prey to better understand the likely predator-prey interactions of P. cygnus. Prey preferences of small sub-legal juvenile lobsters, as well as medium and large legal-sized mature lobsters were investigated using laboratory feeding trials to identify size-associated differences in lobster prey preference. Handling time and diet quality were investigated to estimate energetic cost and gain from consuming different prey which may explain prey choice by lobsters. It was found that large lobsters preferred crabs and mussels while medium and small lobsters preferred crabs over mussels, gastropods, and sea urchins. This suggests that strong predator-prey interactions between P. cygnus and crabs may occur in the wild.

The Unexpected Preference for the fac-Isomer with the Tris(5-ester-substituted-2,2′-bipyridine) Complexes of Ruthenium(II)

The synthesis of a number of new 2,2'-bipyridine ligands, functionalized with bulky ester side groups is reported (L2 - L8). Their reaction with [Ru(DMSO)4Cl2] gives rise to tris-chelate ruthenium(II) metal complexes which show an unusually high proportion of the fac-isomer, as judged by 1H NMR following conversion to the ruthenium(II) complex of 2,2'-bipyridine-5-carboxylic acid methyl ester (L1). The initial reaction appears to have thermodynamic control with the steric bulk of the ligands causing the third ligand to be labile under the reaction conditions used, giving rise to disappointing yields and allowing rearrangement to the more stable facial form. DFT studies indicate that this does not appear to be as a consequence of a metal centered electronic effect. The two isomers of [Ru(L1)3](PF6)2 were separated into the two individual forms using silica preparative plate chromatographic procedures, and the photophysical characteristics of the two forms compared. The results appear to indicate that there is no significant difference in both their room temperature electronic absorption and emission spectra or their excited state lifetimes at 77K.