Ve statistických modelů sémantiky jsou významy slov pouze na základě jejich distribuční vlastnosti.Základní zdroj je zde jeden slovník, který lze použít pro různé úkoly, kde se význam slov reprezentovány jako vektory v vektorového prostoru, a slovní podoby jako vzdálenosti mezi jejich vektorových osobnosti. Pomocí silných podobnosti, může vhodnost podmínek uvedených zejména v souvislosti se vypočítá a používá pro celou řadu úkolů, jeden z nich je slovo smysl Disambiguation. V této práci bylo vyšetřeno několik různých přístupů k modelům z vektorového prostoru a prováděny tak, aby k překročení vyhodnocení vlastního výkonu na Word Sense disambiguation úkolem Prague Dependency Treebank.

Submitted by Rosivalda Pereira (mrs.pereira@ufma.br) on 2017-08-14T20:25:49Z No. of bitstreams: 1 JoaoDallyson.pdf: 6621483 bytes, checksum: 19e928fa3d5789994cc1db5d829e0575 (MD5) / Made available in DSpace on 2017-08-14T20:25:49Z (GMT). No. of bitstreams: 1 JoaoDallyson.pdf: 6621483 bytes, checksum: 19e928fa3d5789994cc1db5d829e0575 (MD5) Previous issue date: 2013-07-05 / Strabismus is a condition that affects approximately 4% of the population causing aesthetic problems, reversible at any age, and irreversible sensory changes that modify the mechanism of vision. The Hirschberg test is one of the types of existing tests to detect such a condition. Detection Systems and computeraided diagnosis are being used with some success in helping health professionals. However, in spite of the increasing routine use of high-tech technologies, the diagnosis and therapy in ophthalmology is not a reality within the strabismus subspecialty. Thus, this thesis aims to present a methodology to detect and automatically diagnose and propose the plan of strabismus surgery through digital images. To do this, the study is organized in seven steps: (1) face segmentation; (2) eye region detection; (3) eyes location; (4) limbus and brilliance location; (5) detection, (6) diagnosis and (7) surgical planning of strabismus. The effectiveness of the study in the indication of the diagnosis and surgical plan was evaluated by the mean diference between the results provided by the methodology and the original indication of the expert. Patients were evaluated for eye positions: PPO, INFRA, SUPRA, DEXTRO and LEVO. The method was 88% accurate in identifying esotropias (ET), 100% in exotropias (XT), 80.33% in hipertropias (HT) and 83.33% in hipotropias (HoT). The overall average error in diagnosis was 5:6 and 3:83 for horizontal and vertical desviations, respectivelly. In planning surgeries of medial rectus muscles the average error was 0.6 mm for recession, and 0.9 mm for ressection. For lateral rectus muscles, the average error was 0.8 mm for recession, and 1 mm for resection. / O estrabismo é uma patologia que afeta cerca de 4% da população, provocando problemas estéticos (reversíveis a qualquer idade) e alterações sensoriais irreversíveis, modi cando o mecanismo da visão. O teste de Hirschberg é um dos tipos de exames existentes para detectar tal patologia. Sistemas de Detecção e Diagnóstico auxiliados por computador estão sendo usados com relativo sucesso no auxílio aos pro fissionais de saúde. No entanto, o emprego rotineiro de recursos de alta tecnologia no auxílio diagnóstico e terapêutico em oftalmologia não é uma realidade dentro da subespecialidade estrabismo. Sendo assim, o presente trabalho tem como objetivo apresentar uma metodologia para detectar e diagnosticar automaticamente, além de propor o plano cirúrgico do estrabismo por meio de imagens digitais. Para tanto, o estudo está organizado em sete estágios: (1) segmentação da face; (2) detecção da região dos olhos; (3) localização dos olhos; (4) localização do limbo e do brilho; (5) detecção; (6) diagnóstico e (7) planejamento cirúrgico do estrabismo. A e ficácia do estudo na indicação do diagnóstico e do plano cirúrgico foi avaliada pela m édia da diferença entre os resultados fornecidos pela metodologia e as indicações originais do especialista. Os pacientes foram avaliados nas posições do olhar: PPO, INFRA, SUPRA, DEXTRO e LEVO. O método obteve acuracia de 88% na identi cação de esotropias (ET), 100% nas exotropias (XT), 80,33% nas hipertropias (HT) e 83,33% nas hipotropias (HoT). O erro médio global na realização do diagnóstico foi de 5:6 e 3:83 para desvios horizontais e verticais, respectivamente. No planejamento de cirurgias de músculos retos mediais obteve-se erro médio de 0,6 mm para recuo, e 0,9 mm para ressecção. Para os músculos retos laterais, o erro médio foi de 0,8 mm para recuo e 1 mm para ressecção.

Les maladies transmises par les moustiques représentent une menace permanente en santé publique. La principale mesure pour protéger les populations contre les piqûres infectieuses repose actuellement sur l’utilisation d’insecticides, mais celle-ci est menacée par la propagation des mécanismes de résistance au sein des populations de moustiques. Dans ce contexte, les répulsifs représentent un outil de choix pour réduire à la fois les nuisances provoquées par les moustiques et le risque d’infection. Parmi eux, le DEET (N, N-diéthyl-3-méthylbenzamide) a prouvé une efficacité remarquable aussi bien lors d’expérimentations en laboratoire que sur le terrain. Malgré cela, les répulsifs ont toujours été utilisés de manière empirique, et leur mode d’action reste très mal connu. Le but de ce projet de thèse a été : i) de participer à l’évaluation du potentiel d’intégration du répulsif DEET dans la lutte antivectorielle, ii) de mesurer l’impact de paramètres physiologiques du moustique d’intérêt épidémiologique tels que l’expérience, l’âge et l’infection sur l’efficacité des méthodes de lutte, répulsifs et insecticides pyréthrinoïdes et iii) de quantifier l’impact des répulsifs et insecticides sur les traits d’histoire de vie des moustiques. Les expériences ont été réalisées sur Anopheles gambiae et Aedes albopictus, respectivement vecteurs de l’agent du paludisme et d’arbovirus tels que le virus de la dengue. Nos résultats ont permis de démontrer que l’état physiologique du moustique influe sur l’efficacité des méthodes de lutte ; premièrement, les moustiques porteurs de la mutation kdr ayant obtenu un repas de sang en contact avec de la perméthrine ne sont plus irrités par ce composé lors d’une seconde exposition, dans les conditions testées et à une dose recommandée pour l’imprégnation de moustiquaires. Au contraire, un repas de sang obtenu en présence DEET n’a pas affecté l’efficacité de ce composé à inhiber un repas de sang à l’exposition suivante. Ensuite, le DEET s’est montré plus efficace chez les moustiques âgés que chez les moustiques jeunes, et les résultats sont similaires chez les deux espèces et indépendants du statut de résistance aux insecticides. D’autre part, l’efficacité de la deltaméthrine et du DEET ne sont pas modifiés lorsque les moustiques sont infectés par le parasite Plasmodium falciparum. Cependant, une augmentation de la mortalité a été observée chez les femelles anophèles infectées au stade sporozoites lors de certaines expérimentations, indépendamment du traitement chimique de la moustiquaire, et suggère un coût de l’infection « réplicat-dépendant ». Enfin, le DEET génère un impact au long terme sur la fécondité et la fertilité des moustiques, effets qui ne sont pas observés dans notre protocole pour la perméthrine. Ces résultats donnent donc des pistes pour redéfinir les priorités dans les programmes de lutte, afin de cibler en premier lieu les moustiques les plus à même de transmettre des pathogènes. Ils soulignent également la nécessité de considérer à la fois l’état physiologique du moustique et l’impact sur le long terme des insecticides et répulsifs lors de l’évaluation des outils de lutte. / Mosquito-borne diseases are a permanent public health threat. The main way to protect populations against infectious bites relies on the use of insecticides, but their efficacy is threatened by the spread of resistance mechanisms among vector populations. In this context, repellents are a promising tool for reducing both mosquito nuisance and infection risk. Among them, DEET (N, N-diéthyl-3-méthylbenzamide) has proven great efficacy both in laboratory and field experiments. Despite this, repellents, are still used from an empirical way and their mode of action remains poorly understood. The aim of the present project was to: i) participate to the evaluation of the potential of integration of the DEET repellent in mosquito control, ii) measure the impact of mosquito physiological parameters of epidemiological relevance such as experience, age and infection, on the efficacy of repellents and insecticides, and iii) quantify the impact of repellents and insecticides on mosquito life-history traits. Experiments were performed on Anopheles gambiae and Aedes albopictus, vectors of the malaria parasite and of arboviruses such as dengue, respectively. Our results allowed to demonstrate that mosquito physiological state influences the efficacy of control tools; first, after successfully obtaining a blood meal in contact with permethrin, mosquitoes carrying kdr mutation are no longer inhibited by this compound at the subsequent exposure, at the recommended dose for net impregnation and under our experimental design. On the contrary, a first blood meal obtained in presence of DEET does not affect its efficacy at the following exposure. Then, DEET appeared to be more efficient against old mosquitoes that against younger ones, and results are consistent in the two species, independently of the insecticide resistance status. Besides, efficacy of DEET and deltamethrin is not modified when mosquitoes are infected with Plasmodium falciparum. However, an increased mortality was observed in anopheles when infected with transmissible stages (i.e. sporozoites), independently of chemical exposure, which suggest a cost of infection and seems to be “replicate-dependent”. Finally, DEET produces a long-term impact on mosquito fecundity and fertility, which is not observed with permethrin under our protocol. These results give insights to redefine the priorities in mosquito control programs, in order to specifically targeting mosquitoes that are the most susceptible to transmit pathogens. These observations also underline the need for considering both mosquito physiological state and the long-term impact of repellents and insecticides during the evaluation of control tools.

Androgens are essential for the maintenance of male health and wellbeing. A disturbance in androgen signalling has been associated with a number of clinically relevant disorders such as cardiovascular disease, diabetes and metabolic disorders as well as infertility. Primarily produced in the testis in males, the actions of androgens are mediated through binding to androgen receptor (AR), a member of the nuclear receptor superfamily of ligand-activated transcription factors. The somatic cells of the testis are known to have a number of key roles in both testis function and development and the Sertoli, Leydig and Peritubular Myoid cells are known to express AR in adulthood. It is through AR that some testicular functions are mediated; for example, the Sertoli cells support of complete spermatogenesis with Sertoli cell androgen receptor knockout (SCARKO) testis demonstrating a halt of spermatogenesis before meiosis. However, how androgen signalling is impacting testicular function through each of the somatic cell types is not yet fully understood. Currently, treatments for male reproductive disorders such as hypogonadism (low androgens) and infertility are limited to treatment of the symptoms; using androgen replacement therapy and in vitro fertilisation techniques. This has been, up until recently, a result of a lack of understanding of the causes of these conditions and a lack of resources able to treat them, with research suggesting that a genetic component may be responsible in a number of cases. However, due to the limited genetic investigation diagnosis of men with male reproductive disorders, the wider understanding of the genetics underpinning male hypogonadism and infertility is incomplete. Developments in technology for the investigation and editing of the genetic code are triggering a surge in the exploration of genetic disorders and, in parallel, into the fields of gene delivery vectors and editing technologies. These technologies will allow an expansion into the knowledge and understanding of genetic disorders whilst simultaneously affording the opportunity to exploit this understanding for the development of therapeutics. There have been a small handful of previous studies using technologies such as viral vectors to target the testicular somatic cells and deliver exogenous transgenes with the purpose of both gene editing and repair, all with varying degrees of success. Here, techniques to introduce and target the Leydig and Sertoli cells were investigated to determine the most appropriate methodology for gene delivery to and manipulation of the testis. Refinement of injections into the interstitial compartment were carried out before introducing lentiviral vectors and targeting of Leydig cells was validated and optimised. Lentiviral vectors are able to permanently integrate into the host cell. Surprisingly, analysis of testis post lentiviral injection determined that the lentiviral targeted Leydig cells began to undergo apoptosis one week post injection and were subsequently cleared from the testis after ten days. Contrastingly, this was not the case when adenoviral vectors were introduced into the interstitial compartment, with Leydig cells continuing to express the delivered reporter transgene and, importantly, not expressing markers of apoptosis, ten days post injection. This would suggest that using adenoviral vectors to target the Leydig cell population in the adult testis would be more appropriate than using lentiviral vectors. Previous studies have successfully used lentiviral vectors to target the Sertoli cells in the adult testis via the introduction of the particles through the efferent duct. However, this can result in damage to efferent duct, resulting in blockages and subsequently the seminiferous tubules. To circumvent this, introduction of the lentiviral particles through the rete compartment of the testis at a range of lower injection pressures was examined and injecting at a lower pressure through the rete testis was found to reduce the likelihood of introducing negative impacts on testicular histology when targeting the seminiferous tubules. Using these refined methods of introducing lentiviral vectors, targeted Sertoli cells stably expressed the delivered transgene for up to one year post injection. Using viral vector delivered transgenes for both the investigation of testicular genetic disorders and for the development of therapeutics has great potential. To explore this potential, we first generated a mouse model in which AR was ablated from both the Leydig and Sertoli cells using Cre/LoxP technology, termed the SC-LC-ARKO. Alongside providing a potential model to 'repair' with viral vectors, the SC-LC-ARKO model also provided an additional model for comparison with other models exhibiting ablation of AR from both single somatic cell types and double somatic cell types. This further enabled a characterisation of the roles of AR in adult testicular function, with results suggesting that loss of AR from more than one cell type results in an additive phenotype when compared to single cell knock outs. Despite providing further insight into the roles of AR in the testis, further analysis of the Cre line used to generate the SC-LC-ARKO model indicated that a small number of Leydig cells were expressing the Cre recombinase, resulting in only a small population of Leydig cells with ablated AR. Considering this, to explore the potential of rescuing Sertoli cell AR using lentiviral vectors, we then utilised an already well characterised Sertoli Cell AR knockout (SCARKO) model. Lentiviral vectors expressing mouse AR and monomeric GFP (moeGFP) downstream of a CMV promoter were generated and injected into the rete testis of WT and SCARKO adult (day 100) males at low pressure. The contralateral testis was injected with a lentiviral vector expressing moeGFP alone (also downstream of a CMV promoter) using the same technique. Analysis of testis sections revealed a reintroduction of AR to Sertoli cells in 100% of SCARKO testis injected with lentivirus expressing mouse AR. As a result of this re-expression of AR in Sertoli cells, 66% of the testis injected with lentivirus expressing mouse AR had evidence of morphologically mature elongated spermatids, indicative of ongoing spermatogenesis. These results suggest that a rescue of the infertility phenotype reported in previous studies of SCARKO testis. Also demonstrated is the reversal of the SCARKO testicular phenotype in tubules targeted by the mAR expressing lentiviral vector. This suggests that absence Sertoli cell AR throughout development does not have a permanent impact on the Sertoli cells capacity to support spermatogenesis in adulthood following rescue of SC AR expression in adulthood. In summary, the results of these studies have provided a refinement in the methodologies for targeting the Sertoli and Leydig cells of the adult testis with viral vectors as well as demonstrating successful rescue of a previously reported mouse model exhibiting infertility through reintroduction of a functional gene. Alongside this, comparisons of AR knockout models have afforded insight into maintenance of testis function through AR.

The reliability of monetary policy as an economic stabilisation tool depends on the understanding of the empirical effects of policy intervention on macroeconomic aggregates. Since investigating the interdependencies between macroeconomic variables necessarily involves studying their interactions over time, time series analysis is an important tool. This thesis sets out to examine the presence and effects of nonstationarity in the form of a structural break in a basic VAR of four endogenous variables. Specifically, the transmission of a monetary policy shock on the macroeconomic aggregate of 11 Euro Area countries is estimated for the period 1999–2017, employing variables based on previous studies. A Quandt-Andrews breakpoint test is used to identify the break date, and a comparison is made between the periods. This study finds support for the presence of a break in the regression estimate from the breakpoint test, although the reults from the IRFs cannot be shown to be statistically significant, nor to be bias-free.

Tsoi, Hung Ming. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2008. / Includes bibliographical references (leaves 88-90). / Abstracts in English and Chinese. / Preface --- p.7 / Chapter 1 --- Kahler Manifold --- p.10 / Chapter 1.1 --- Hermitian Manifold --- p.12 / Chapter 1.2 --- Kahler Manifold --- p.13 / Chapter 1.2.1 --- "Positive (l,l)-form" --- p.15 / Chapter 2 --- Vector Bundle --- p.16 / Chapter 2.1 --- Holomorphic Vector Bundle and Connection --- p.17 / Chapter 2.2 --- Hermitian Connection and Chern Connection --- p.18 / Chapter 2.2.1 --- Existence of Chern connection on a holomorphic vector bundle --- p.19 / Chapter 2.3 --- Curvature --- p.21 / Chapter 2.4 --- Positivity of Vector Bundles --- p.23 / Chapter 2.5 --- Chern Classes and Holomorphic Line Bundle --- p.24 / Chapter 2.5.1 --- Chern class in axiomatic approach --- p.25 / Chapter 2.5.2 --- Chern class in algebraic topology --- p.26 / Chapter 2.5.3 --- Chern class in terms of curvature --- p.27 / Chapter 2.5.4 --- In the case of hermitian line bundle --- p.28 / Chapter 3 --- Analytic Technique on Kahler Manifold --- p.30 / Chapter 3.1 --- Dolbeault Cohomology --- p.30 / Chapter 3.2 --- Commutator Relations on Kahler Manifold --- p.31 / Chapter 3.2.1 --- Commutator relation on a line bundle --- p.32 / Chapter 3.3 --- Hodge Theory --- p.33 / Chapter 3.4 --- Bochner Technique --- p.35 / Chapter 3.4.1 --- Bochner-Kodaira-Nakano identity --- p.36 / Chapter 4 --- Kodaira Vanishing Theorem and L2 estimate of d --- p.38 / Chapter 4.1 --- Kodaira Vanishing Theorem --- p.39 / Chapter 4.2 --- Extension of Kodaira Vanishing Theorem by L2 Method --- p.44 / Chapter 4.2.1 --- Plurisubharmonic functions and weakly pseudoconvex Kahler manifold --- p.47 / Chapter 5 --- Multiplier Ideal Sheaf --- p.55 / Chapter 5.1 --- Algebraic Properties of Multiplier Ideal Sheaf --- p.56 / Chapter 5.2 --- Some Calculations of Multiplier Ideal Sheaf --- p.59 / Chapter 6 --- Nadel Vanishing Theorem --- p.62 / Chapter 6.1 --- Nadel Vanishing Theorem by L2 Estimate of d --- p.62 / Chapter 6.2 --- The Original Setting of Nadel --- p.64 / Chapter 6.2.1 --- S-bounded and S-null sequence --- p.65 / Chapter 6.2.2 --- Multiplier ideal sheaf by Nadel --- p.67 / Chapter 6.3 --- Nadel Vanishing Theorem by Computation of Cech Cohomology --- p.69 / Chapter 6.3.1 --- L2 estimate of d --- p.69 / Chapter 6.3.2 --- Koszul cochain --- p.70 / Chapter 6.3.3 --- The cohomology vanishing theorem --- p.73 / Chapter 7 --- Kawamata-Viehweg Vanishing Theorem --- p.77 / Chapter 7.1 --- Numerically Effective Line Bundle --- p.77 / Chapter 7.2 --- Kawamata-Viehweg Vanishing Theorem --- p.85 / Bibliography --- p.88

This work offers two strategies to raise the prediction accuracy of Vector Autoregressive (VAR) Models. The first strategy is to improve the Minnesota prior, which is frequently used for Bayesian VAR models. The improvement is achieved in two ways. First, the variance-covariance matrix of regression disturbances is treated as unknown and random to incorporate parameter uncertainty. Second, the prior variance-covariance matrix of regression coefficients is constructed as a function of the variance-covariance matrix of disturbances, in order to account for dependencies between different equations. Since different prior specifications unavoidably lead to different models, and forecasting capability of any such model is often limited, the second strategy is to build an optimal prediction pool of models by using the conventional log predictive score function. The effectiveness of the proposed strategies is examined for one-step-ahead, multi-4-step-ahead, and single-4-step-ahead predictions through two exercises. One exercise is predicting national output, inflation, and interest rate in the United States, and the other is predicting state tax revenue and personal income in Iowa. The empirical results indicate that a properly selected prior can improve the prediction performance of a BVAR model, and that a real-time optimal prediction pool can outperform a single best constituent model alone.

Navigating through the debris of the information explosion requires powerful, flexible search tools. These tools must be both useful and useable; that is, they must do their jobs effectively without placing too many burdens on the user. While general interest search engines, such as Google, have addressed this latter challenge well, more topic-specific search engines, such as PubMed, have not. These search engines, though effective, often require training in their use, as well as in-depth knowledge of the domain over which they operate. Furthermore, search results are often returned in an order irrespective of users' preferences, forcing them to manually search through search results in order to find the documents they find most useful. To solve these problems, we intend to learn ranking functions from user relevance preferences. Applying these ranking functions to search results allows us to improve search usability without having to reengineer existing, effective search engines. Using ranking SVMs and active learning techniques, we can effectively learn what is relevant to a user from relatively small amounts of preference data, and apply these learned models as ranking functions. This gives users the convenience of seeing relevance-ordered search results, which are tailored to their preferences as opposed to using a one-size-fits-all sorting method. As giving preference feedback does not require in-depth domain knowledge, this approach is suitable for use by domain experts as well as neophytes. Furthermore, giving preference feedback does not require a great deal of training, adding very little overhead to the search process.

Cystic fibrosis (CF) is an autosomal recessive genetic disorder of which lung disease is the leading cause of morbidity and mortality. One attractive strategy for the treatment of CF lung disease is to directly deliver CF transmembrane conductance regulator gene to airway epithelia. Although promising results have been reported, barriers present in the lung make successful gene transfer to the respiratory tract difficult. In order to improve gene transfer strategies in the intrapulmonary airways, we need to identify the bottlenecks of transduction for the vector system. A previous study reported that feline immunodeficiency virus (FIV)-mediated gene transfer was more efficient in the nasal airways in mice than the intrapulmonary airways (Sinn, P.L. et al. 2008, J. Viol). Our first goal was to identify barriers to lentiviral gene transfer in the murine airways. We demonstrate that host immune response is not the major barrier preventing efficient FIV-mediated transduction in the intrapulmonary airways. We show that the FIV vector transduces murine primary nasal epithelial cell cultures with greater efficiency than murine primary tracheal epithelial cell cultures. In addition, GP64 pseudotyped vesicular stomatitis virus (VSV) transduces better in nasal epithelia compared to intrapulmonary airways in mice. On the other hand, we observed that VSVG glycoprotein-pseudotyped VSV transduces the intrapulmonary airway as well as nasal epithelia in mice with similar efficiency. Our results suggest that differentially expressed cellular factor(s) specific for GP64 or FIV vector may be the major barrier(s) for FIV vector-mediated gene transfer in the murine intrapulmonary airways. The recent development of CF porcine models prompted us to investigate possible barriers for lentiviral vector-mediated gene transfer in porcine cells. Our preliminary results showed that HIV transduction was restricted in porcine but not human lung-derived cell lines. Porcine TRIM5 has sequences similar to restrictive bovine TRIM5 orthologs. Therefore, our second goal was to investigate the possible restriction of lentiviral vectors by porcine TRIM5. We demonstrate that transient overexpression or knockdown of endogenously expressed porcine TRIM5 does not affect HIV or FIV transduction. Lastly, we characterized a mucin domain-deleted EBOV (EBOVΔO) glycoprotein mutant with increased transduction. This EBOVΔO 5-mer mutant was generated based on mutants with an increased transduction as identified by alanine scanning mutagenesis (Brindlay, M.A. et al. 2007. J. Viol). We show that VSV pseudotyped with the 5-mer mutant increased transduction both in vitro and in mice when compared to the wild-type EBOVΔO. Structural analysis demonstrated that 5 mutations were located proximal to the GP1-GP2 interface. Enhanced transduction likely results from a lower energy metastable state of the glycoprotein. FIV pseudotyped with 5-mer also shows increased transduction in multiple cell lines. Identification of barriers in intrapulmonary airways and improvements of vector systems will help the advancement of gene therapy for CF.

Neutrosophic logic is a relatively new logic that is a generalization of fuzzy logic. In this dissertation, for the first time, neutrosophic logic is applied to the field of classifiers where a support vector machine (SVM) is adopted as the example to validate the feasibility and effectiveness of neutrosophic logic. The proposed neutrosophic set is integrated into a reformulated SVM, and the performance of the achieved classifier N-SVM is evaluated under an image categorization system. Image categorization is an important yet challenging research topic in computer vision. In this dissertation, images are first segmented by a hierarchical two-stage self organizing map (HSOM), using color and texture features. A novel approach is proposed to select the training samples of HSOM based on homogeneity properties. A diverse density support vector machine (DD-SVM) framework that extends the multiple-instance learning (MIL) technique is then applied to the image categorization problem by viewing an image as a bag of instances corresponding to the regions obtained from the image segmentation. Using the instance prototype, every bag is mapped to a point in the new bag space, and the categorization is transformed to a classification problem. Then, the proposed N-SVM based on the neutrosophic set is used as the classifier in the new bag space. N-SVM treats samples differently according to the weighting function, and it helps reduce the effects of outliers. Experimental results on a COREL dataset of 1000 general purpose images and a Caltech 101 dataset of 9000 images demonstrate the validity and effectiveness of the proposed method.