One of the most active research areas in computer vision is image classification. Although there have been many research efforts in this area, it remains a difficult problem, especially when the number of categories is large. Most of the previous work in image classification uses low-level image features. We believe low-level features ignore a lot of the semantic structures of the image classes. In this thesis, we go beyond simple low-level features and propose new approaches for constructing mid-level visual features for image classification. We represent an image using the outputs of a collection of binary classifiers. These binary classifiers are trained to differentiate pairs of object classes in an object hierarchy. Our feature representations implicitly capture the hierarchical structure in object classes. We show that our proposed approach outperforms other baseline methods in image classification.

Knee replacement surgery is a common orthopaedic procedure that greatly benefits from a three-dimensional geometric representation of a patient's knee bone obtained from MR or CT data. The use of these image modalities pose the following challenges: (i) high imaging cost; (ii) long wait times; (iii) limited availability and (iv) in the latter, large exposure to ionizing radiation. Traditional approaches based on planar X-ray radiography are significantly less prone to these issues; however, they only provide two-dimensional information. This work presents a proof of concept study for generating patient-specific femoral bone shapes from a statistical shape atlas using anatomical features acquired from calibrated X-ray radiographs. Our hypothesis was: three-dimensional geometry can be reconstructed, within 2 millimeters RMS, by identifying features on two calibrated radiographs. We illustrate the feasibility of our approach with regards to acquiring features and the viability of reconstructing patient-specific bony anatomy. A set of reliable and relevant features is identified for which an acquisition protocol and user-interface was devised to minimize inter-observer variability. Both the data and methods used to construct the atlas are discussed as well generating shapes from features. The reconstructions accuracy was comparable to, albeit lower than, competing approaches that rely on two-dimensional bony contours. / Thesis (Master, Computing) -- Queen's University, 2014-04-29 21:53:10.809

Deciding which features to include on a sprayer in order to increase a purchaser's likelihood of buying was a major problem for spray equipment manufacturers. There were several sprayer options that could be included or excluded that affect the retail price of the cotton sprayer. Conjoint analysis was utilized to determine the relative rank and value of features on a cotton sprayer. At the Georgia Cotton Commission Conference, ninety-five respondents completed a sorting of eight cards that each contained groupings of five cotton sprayer feature options. A demographics page was used to identify respondents that were farmer-buyers. The number of participants used in the study was fifty-six. The rankings from this study were entered into SPSS statistical software to retrieve utility values, importance values, and correlations. The analysis of the data showed that the inclusion of chlorophyll sensors had the largest influence on a purchaser's decision to buy. Following this, in terms of importance was presence of wheel shields, type of wheel tread adjustment, and number of spray boom sections. The highest ranking feature combination was three chlorophyll sensors, the presence of shields, hydraulic tread adjustment, and two boom plumbing sections. Chlorophyll sensors were twice as important to respondents as all other features. The respondents were willing to pay the extra cost for the three chlorophyll sensors, proving that the technology was important to them. Once the price increased additionally for the six chlorophyll sensors the respondents' preference for the technology was overshadowed by their preference for price. This showed a strong trade-off with price. It appears that they may be unwilling to pay for the technology because they do not fully understand the benefits of variable rate technology or feel that the technoology cost will not be offset with the benefits.

Through humoral and cell-mediated mechanisms, the immune system plays a vital role in protecting every organ system. Disorders of the immune system may result in various cutaneous manifestations, including cutaneous malignancies. In patients with immunodeficiency, the risk of development of malignant cutaneous neoplasms is substantially increased. This increased risk may be due to oncogenic viruses that find a suitable microenvironment for tumorigenesis and cancer development. A subset of cutaneous malignancies that develop in patients with immunodeficiency may show aggressive clinical and biological behavior. Here, we report six cases of highly aggressive and deadly cutaneous neoplasms that arose in patients with a known history of immunodeficiency: two cases of Kaposi sarcoma in patients with immunosuppression due to human immunodeficiency virus infection; a case of Merkel cell carcinoma and a case of squamous cell carcinoma (SCC) in patients receiving immunosuppressive drugs after organ transplant; a case of multiple cutaneous tumors, including invasive melanoma, SCC, and sebaceous carcinoma, in a patient with hypogammaglobulinemia and a history of organ transplant; and a case of basal cell carcinoma and melanoma in situ in a patient with primary immunodeficiency.

<p> When pigeons are required to discriminate between two displays which may only be differentiated by a distinctive feature on one of the two displays, subjects trained with the distinctive feature on the positive display learn the successive discrimination while subjects trained with the distinctive feature on the negative display do not. The simultaneous discrimination theory of this "feature-positive effect" makes a number of explicit predictions about the behaviour of the feature positive and feature negative subjects. The present experiments were designed to test these predictions. Experiment I tested the prediction of localization on the distinctive feature by feature positive subjects while Experiment II tested the prediction of avoidance of the distinctive feature by feature negative subjects. Experiment III attempted to reduce the feature-positive effect by presenting compact displays.</p> <p> The results of these three experiments supported the simultaneous discrimination theory of the feature positive effect.</p> / Thesis / Doctor of Philosophy (PhD)

Constitutions and bills of rights have previously been argued to be non-democratic. To justify the entrenched nature of constitutions, some theorists have argued that constitutions represent a type of rational precommitment. However, this precommitment understanding of constitutions is not without its own problems. In this work, I will argue the prominent understanding of constitutional precommitment used by its proponents seems to rely upon a definition of commitment to which their arguments do not stay true. However, when I try to amend their arguments and apply a proper example of commitment, it leads to some problems with other tenets of the constitutional debate, especially the fact of constitutional entrenchment. In an attempt to determine just what it would take to save the rational precommitment understanding of constitutions, while maintaining a proper definition of commitment, I turn to metaphysical puzzles about change, persistence, and the possibility of a mereological understanding of our constitution. I conclude that 1) current debates do not have a proper conception of commitment and are thus failing to accomplish their ends, and 2) if proponents of the rational precommitment view do not buy into my analysis, then it is going to prove quite difficult to keep their account afloat once we properly define commitment. / Thesis / Master of Arts (MA)

A huge amount of data is generated daily by digital technologies such as social media, web logs, traffic sensors, on-line transactions, tracking data, videos, and so on. This has led to the archiving and storage of larger and larger datasets, many of which are multi-modal, or contain different types of data which contribute to the problem that is now known as “Big Data”. In the area of Big Data, volume, variety and velocity problems remain difficult to solve. The work presented in this thesis focuses on the variety aspect of Big Data. For example, data can come in various and mixed formats for the same feature(attribute) or different features and can be identified mainly by one of the following data types: real-valued, crisp and linguistic values. The increasing variety and ambiguity of such data are particularly challenging to process and to build accurate machine learning models. Therefore, data heterogeneity requires new methods of analysis and modelling techniques to enable useful information extraction and the modelling of achievable tasks. In this thesis, new approaches are proposed for handling heterogeneous Big Data. these include two techniques for filtering heterogeneous data objects are proposed. The two techniques called Two-Dimensional Similarity Space(2DSS) for data described by numeric and categorical features, and Three-Dimensional Similarity Space(3DSS) for real-valued, crisp and linguistic data are proposed for filtering such data. Both filtering techniques are used in this research to reduce the noise from the initial dataset and make the dataset more homogeneous. Furthermore, a new similarity measure based on intuitionistic fuzzy set theory is proposed. The proposed measure is used to handle the heterogeneity and ambiguity within crisp and linguistic data. In addition, new combine similarity models are proposed which allow for a comparison between the heterogeneous data objects represented by a combination of crisp and linguistic values. Diverse examples are used to illustrate and discuss the efficiency of the proposed similarity models. The thesis also presents modification of the k-Nearest Neighbour classifier, called k-Nearest Neighbour Weighted Average (k-NNWA), to classify the heterogeneous dataset described by real-valued, crisp and linguistic data. Finally, the thesis also introduces a novel classification model, called FCCM (Filter Combined Classification Model), for heterogeneous data classification. The proposed model combines the advantages of the 3DSS and k-NNWA classifier and outperforms the latter algorithm. All the proposed models and techniques have been applied to weather datasets and evaluated using accuracy, Fscore and ROC area measures. The experiments revealed that the proposed filtering techniques are an efficient approach for removing noise from heterogeneous data and improving the performance of classification models. Moreover, the experiments showed that the proposed similarity measure for intuitionistic fuzzy data is capable of handling the fuzziness of heterogeneous data and the intuitionistic fuzzy set theory offers some promise in solving some Big Data problems by handling the uncertainties, and the heterogeneity of the data.

With roughly about half of the energy used in buildings attributed to Heating, Ventilation, and Air conditioning (HVAC) systems, there is clearly great potential for energy saving through improved building operations. Accurate knowledge of localised and real-time occupancy numbers can have compelling control applications for HVAC systems. However, existing technologies applied for building occupancy measurements are limited, such that a precise and reliable occupant count is difficult to obtain. For example, passive infrared (PIR) sensors commonly used for occupancy sensing in lighting control applications cannot differentiate between occupants grouped together, video sensing is often limited by privacy concerns, atmospheric gas sensors (such as CO2 sensors) may be affected by the presence of electromagnetic (EMI) interference, and may not show clear links between occupancy and sensor values. Past studies have indicated the need for a heterogeneous multi-sensory fusion approach for occupancy detection to address the short-comings of existing occupancy detection systems. The aim of this research is to develop an advanced instrumentation strategy to monitor occupancy levels in non-domestic buildings, whilst facilitating the lowering of energy use and also maintaining an acceptable indoor climate. Accordingly, a novel multi-sensor based approach for occupancy detection in open-plan office spaces is proposed. The approach combined information from various low-cost and non-intrusive indoor environmental sensors, with the aim to merge advantages of various sensors, whilst minimising their weaknesses. The proposed approach offered the potential for explicit information indicating occupancy levels to be captured. The proposed occupancy monitoring strategy has two main components; hardware system implementation and data processing. The hardware system implementation included a custom made sound sensor and refinement of CO2 sensors for EMI mitigation. Two test beds were designed and implemented for supporting the research studies, including proof-of-concept, and experimental studies. Data processing was carried out in several stages with the ultimate goal being to detect occupancy levels. Firstly, interested features were extracted from all sensory data collected, and then a symmetrical uncertainty analysis was applied to determine the predictive strength of individual sensor features. Thirdly, a candidate features subset was determined using a genetic based search. Finally, a back-propagation neural network model was adopted to fuse candidate multi-sensory features for estimation of occupancy levels. Several test cases were implemented to demonstrate and evaluate the effectiveness and feasibility of the proposed occupancy detection approach. Results have shown the potential of the proposed heterogeneous multi-sensor fusion based approach as an advanced strategy for the development of reliable occupancy detection systems in open-plan office buildings, which can be capable of facilitating improved control of building services. In summary, the proposed approach has the potential to: (1) Detect occupancy levels with an accuracy reaching 84.59% during occupied instances (2) capable of maintaining average occupancy detection accuracy of 61.01%, in the event of sensor failure or drop-off (such as CO2 sensors drop-off), (3) capable of utilising just sound and motion sensors for occupancy levels monitoring in a naturally ventilated space, (4) capable of facilitating potential daily energy savings reaching 53%, if implemented for occupancy-driven ventilation control.

The object of study of this thesis is the linguistic input in textbooks of Italian as a foreign language (FL). The intent is to study whether the linguistic changes, observed in contemporary Italian, have become part of the Italian offered as input to the learners. To identify the variety of language presented in the textbooks, some features of contemporary linguistic changes were chosen as verifiable indicators. These features, listed by Sabatini (1985: 155) as a basic part of "italiano dell’uso medio", and by Berruto (1987: 62) as part of "neostandard", are not occasional changes but are features that are gradually expanding and stabilizing into Italian standard (Sobrero 2005). A corpus consisting of 38 Italian textbooks published in Sweden and 8 in Italy in the years 2000-2012 were used to verify the manifestation of these features. The results show that the presence of neostandard features in the textbooks of Italian FL is conditioned, at first, by the rate of acceptance of those features by the linguistic norm. Thus, features that are nowadays commonly considered as normative have a high number of occurrences in the corpus. This is the case concerning lui, lei, loro as subject pronouns, the use of gli instead of loro, the use of the present tense for the future and the use of temporal che. On the other hand, features that are not considered as normative have no or very few occurrences. This is the case with gli instead of le and the use of imperfetto ipotetico. Secondly, the presence of the neostandard features in textbooks is conditioned by the instructive function of the textbooks, which shapes the typology of input introduced. Thus, occurrences of features such as cleft clauses and dislocations are mainly presented in authentic texts, oral texts, or introduced explicitly, but are rare or absent in textbooks characterized by simplified language.

This study analyzed instructional achievement and intrinsic motivation among 21st century secondary students utilizing a video lecture incorporating both student reaction cutaway images and immediate content interaction within the lecture. Respondents (n = 155) were from multiple classes and grade levels at a suburban Texas high school. Four groups of students viewed the identical lecture with differing video and content interaction treatments. Students responded to a pretest/posttest survey to assess academic achievement in addition to an intrinsic motivation instrument to assess student interest. Group one (the control group) viewed the 12 minute lecture without enhancement. A second group viewed the identical lecture with student reaction shots inserted in the video. Another group viewed the lecture with content question intervention inserted into the video. The final group saw the lecture with the student reaction shots and content question intervention combined in the video. A repeated measures multivariate analysis of variance (MANOVA) was used to compare results from a 14 item pretest/posttest. Combined, the groups showed no significance (p = .069) indicating no associations were identified by the experiment. Although no association was identified, this may be a reflection of the generic nature of the video lecture and the lack of association with the experiment and actual classroom content within their courses. Students also completed the Intrinsic Motivation Instrument which was analyzed using a MANOVA. Although no significant findings were present in either group viewing the student reaction or the content question interaction treatments individually, the group viewing the combined treatment showed significance in three scales: Interest/Enjoyment (p = .007), Perceived Competence (p = .027) and Effort/Importance (p = .035) Recommendations for refinement of the current experiment as well as future studies are provided.