The goal of the diploma thesis is to design, implement and evaluate classifiers for automatic classification of semantic patterns of English verbs according to a pattern lexicon that draws on the Corpus Pattern Analysis. We use a pilot collection of 30 sample English verbs as training and test data sets. We employ standard methods of machine learning. In our experiments we use decision trees, k-nearest neighbourghs (kNN), support vector machines (SVM) and Adaboost algorithms. Among other things we concentrate on feature design and selection. We experiment with both morpho-syntactic and semantic features. Our results show that the morpho-syntactic features are the most important for statistically-driven semantic disambiguation. Nevertheless, for some verbs the use of semantic features plays an important role.

A neuroimaging feature extraction model is designed to extract region-based image features whose values are predicted by base learners trained on raw neuroimaging morphological variables. The main objectives are to identify Alzheimer’s disease (AD) in its earliest manifestations, and be able to predict and gauge progression of the disease through the stages of mild cognitive impairment (EMCI), late MCI (LMCI) and AD. The model was evaluated on the ADNI database and showed 75.26% accuracy for the challenging EMCI diagnosis based on the 10-fold cross-validation. Our approach also performed well for the other binary classifications: EMCI vs. LMCI (72.3%), EMCI vs. AD (95%), LMCI vs. AD (84.3%), CN vs. LMCI (77.5%), and CN vs. AD (96.5%). By applying the model to the Genome-wide Association Study, along with the sparse Partial Least Squares regression method, we successfully detected risk genes such as the APOE, TOMM40, RVRL2 and APOC1 along with the new finding of rs917100. Moreover, the research aimed to investigate the relationship of different biomarkers; especially the imaging biomarkers to better understand the precise biologic changes that characterize Alzheimer’s disease. The unique and independent contribution of APOE4 allele status (E4+\E4-), amyloid (Aβ) load status (Amy+\Amy-) and combined APOE4 and Aβ status on regional cortical thickness (CTh) and cognition were evaluated via a series of two-way ANCOVAs with post-hoc Tukey HSD tests. Results showed that decreased CTh is independently associated with Amy+ status in many brain regions, but with E4+ status in very restricted number of brain regions. Among CN and EMCI participants, E4+ status is associated with increased CTh, in medial and inferior temporal regions. Diverging association patterns of global and regional Aβ load with cortical volume were found in the entorhinal, temporal pole and parahippocampal regions, which were positively associated with regional Aβ load, but with a negative correlation for global Aβ load in MCI stages. In addition, strong positive correlations were shown between baseline regional CTh and the difference of CTh in each region between the CN and AD, even after adjusting for the regional Aβ and APOE genotype (E4+: r = 0.521 and E4-: r = 0.694).

Lightning-caused forest fires are one of the major natural disturbances in Ontario managed boreal forests. Survival of these forests with fires for centuries shows that such disturbances are integral to the boreal ecosystem and its ecological functioning. Characterizing the fire regimes defined by fire ignition frequency, fire sizes and their spatial distribution patterns etc. thus can help to improve our understanding of the boreal forest dynamics and provide guidance for management practices attempting to maintain biodiversity and achieve sustainability. In this thesis the lightning-caused fire ignitions data for four ecoregions in Ontario managed boreal forests (3E, 3W, 3S and 4S) for 1960–2009 were analyzed using pattern analysis and density estimation to determine the spatial nature of fire ignitions. These fire ignition spatial patterns were further used (as weighted ignition scenario) to simulate forest fire regimes in the study area. Fire regimes were also simulated using spatially unweighted ignitions (unweighted ignition scenario). Non-spatial (total number of fires, total burn area, number of fires by size classes, annual burn fraction) and spatial (spatial burn probability) indicators of the simulated fire regimes under both ignition scenarios were compared to test the null hypothesis that modeled forest fire regime is not affected by the spatial patterns of input fire ignitions. All data analysis were performed for individual ecoregions. Spatial pattern of ignitions were analyzed using the nearest neighbour index and Ripley’s K-function. Ignition densities were estimated using the adaptive kernel density estimation method and the fire regimes were simulated using BFOLDS (Boreal Forests Landscape Dynamics Simulator). Results showed that lightning-caused fire ignitions are clustered in all ecoregions. Fire ignition density also varied spatially within ecoregions. Overall fire ignition density was highest in the northwestern ecoregion (4S) and lowest in the eastern ecoregion (3E), which corresponds to the combined gradient of effective humidity and temperature in Ontario. For each ecoregion, comparison of non-spatial simulated fire regime indicators showed statistically non-significant differences between unweighted and weighted ignitions. The spatial burn probability however captured clear spatial differences between unweighted and weighted ignitions. Spatial differences in spatial burn probability between both ignition scenarios were more prominent in ecoregions of high fire occurrence. Results of the weighted ignition scenario closely followed the spatial patterns of the estimated fire ignition density in the study area. Based on these results this thesis rejects the null hypothesis and emphasizes that ignition patterns must be considered in simulating fire regime in Ontario boreal forests.

With increasing levels of motor vehicle ownership, automobile crashes have become a serious public issue in the U.S. and around the world. Knowing when, where, and how traffic accidents happen is critical in order to ensure road safety and to plan for adequate road infrastructure. There is a rich body of literature pertaining to time-related fatal crashes, most of which focuses on non-spatial factors such as a driver’s visibility at night, drinking and drug use, and road conditions. These studies provide a theoretical basis for understanding the causes of crashes from a non-spatial perspective, and a number of traffic laws and policies consequently have been enacted to minimize the impacts of non-spatial factors. Over the past few years, advances in Geographic Information Systems (GIS) have greatly enhanced our ability to analyze traffic accidents from a spatial perspective. This study aims to fill a void in traffic safety studies by comparing and analyzing the differences in the spatial distribution of fatal crashes based on temporal factors, specifically in three periods: 1) day and night; 2) A.M. rush hours and P.M. rush hours; and 3) weekdays and weekends. With the Nashville Metropolitan Area as the study area, the research utilized a number of spatial point-pattern analysis (SPPA) methods, including planar KDE, planar global auto K function, network global cross K functions, and network local cross K functions. All fatal crashes in the Nashville area were found to be clustered and generally follow the patterns of average daily traffic flow. All time-based subtypes of fatal crashes also were found to be concentrated within the central urban area of Nashville, mostly along major roads, and especially near major road intersections and highway interchanges. No notable spatial differences were detected among the subtypes of fatal crashes when applying network global cross K function. However, with the help of the network local cross K function, some localized spatial differences were identified. Some specific locations of hotspots of nighttime and P.M. rush hour fatal crashes were found not to be at the same locations as those at of daytime and A.M. rush hour fatal crashes, respectively. The approach adopted in this study not only provides a new way to analyze spatial distribution of spatial point events such as fatal crashes, but it also can be applied readily to real-world applications. A good understanding of where these spatial differences are should help various agencies practice effective measures and policies in order to improve road conditions, reduce traffic accidents, and ensure road safety.

Lightning-caused forest fires are one of the major natural disturbances in Ontario managed boreal forests. Survival of these forests with fires for centuries shows that such disturbances are integral to the boreal ecosystem and its ecological functioning. Characterizing the fire regimes defined by fire ignition frequency, fire sizes and their spatial distribution patterns etc. thus can help to improve our understanding of the boreal forest dynamics and provide guidance for management practices attempting to maintain biodiversity and achieve sustainability. In this thesis the lightning-caused fire ignitions data for four ecoregions in Ontario managed boreal forests (3E, 3W, 3S and 4S) for 1960–2009 were analyzed using pattern analysis and density estimation to determine the spatial nature of fire ignitions. These fire ignition spatial patterns were further used (as weighted ignition scenario) to simulate forest fire regimes in the study area. Fire regimes were also simulated using spatially unweighted ignitions (unweighted ignition scenario). Non-spatial (total number of fires, total burn area, number of fires by size classes, annual burn fraction) and spatial (spatial burn probability) indicators of the simulated fire regimes under both ignition scenarios were compared to test the null hypothesis that modeled forest fire regime is not affected by the spatial patterns of input fire ignitions. All data analysis were performed for individual ecoregions. Spatial pattern of ignitions were analyzed using the nearest neighbour index and Ripley’s K-function. Ignition densities were estimated using the adaptive kernel density estimation method and the fire regimes were simulated using BFOLDS (Boreal Forests Landscape Dynamics Simulator). Results showed that lightning-caused fire ignitions are clustered in all ecoregions. Fire ignition density also varied spatially within ecoregions. Overall fire ignition density was highest in the northwestern ecoregion (4S) and lowest in the eastern ecoregion (3E), which corresponds to the combined gradient of effective humidity and temperature in Ontario. For each ecoregion, comparison of non-spatial simulated fire regime indicators showed statistically non-significant differences between unweighted and weighted ignitions. The spatial burn probability however captured clear spatial differences between unweighted and weighted ignitions. Spatial differences in spatial burn probability between both ignition scenarios were more prominent in ecoregions of high fire occurrence. Results of the weighted ignition scenario closely followed the spatial patterns of the estimated fire ignition density in the study area. Based on these results this thesis rejects the null hypothesis and emphasizes that ignition patterns must be considered in simulating fire regime in Ontario boreal forests.

The causes of errors in the angle of impact calculation were investigated including the surface type, falling velocity and the method used to fit an ellipse to a bloodstain. As had been cited previously the angle of impact was generally underestimated, especially at acute angles and the reason for this was determined to be due to an overestimation of the length of a bloodstain. The surface type was found to significantly affect the accuracy of an angle of impact calculation and as the falling velocity increased, the angle of impact calculation became more accurate. High-speed photography was used to further investigate the formation of bloodstains on surfaces. It was found that the formation of the bloodstain varied depending on the surface type and the angle of the surface. Bloodstain pattern analysis involves the application of scientific techniques to reconstruct events that resulted in a bloodstain pattern. The position of the blood source in three-dimensional space is a fundamental element of this application. Currently little is known about the methods used by bloodstain pattern analysts to select bloodstains when determining the region of origin. Fourteen analysts worldwide were surveyed in order to ascertain this information. It was found that the methods used were variable and were often not based on scientific research. Research was therefore undertaken into bloodstain selection and in particular, which bloodstains should be selected for a region of origin analysis. As a result of these experiments, two sets of selection criteria were established, one for use when the region of origin is being calculated manually and one for when directional analysis is being used.

Thesis (MSC (Mathematical Sciences. Computer Science))--Stellenbosch University, 2008. / This thesis describes the design and implementation of an application names GrailKnights that allows for the mass manipulation of automata, with added visual pattern analysis features. It comprises a database-driven backend for automata storage, and a graphical user interface that allows for filtering the automata selected from the database with visual interpretation of visible patterns over the resulting automata.

No description available.

Pour de nombreux écologistes, les forêts tropicales sont synonymes de richesse et de diversité spécifique ; aussi l'existence de forêts monodominantes, à savoir de forêt dominée par une seule espèce, reste un formidable énigme. Nous présentons ici la première étude de l'autécologie d'une nouvelle espèce monodominante : Spirotropis longifolia (DC) Baill. (Leguminosae-Papilionoideae) se développant en Guyane française. La monodominance du S. longifolia est très importante, ce dernier pouvant représenter jusqu'à 70% du peuplement. Par ailleurs, le cortège floristique associé à cette espèce diffère largement de celui de la forêt adjacente, très diversifiée. La monodominance du S. longifolia ne peut être expliquée ni par des conditions pédologiques particulières ni par l'absence de compétiteur dans ses peuplements. Nous suggérons en revanche que ses étonnantes capacités à réitérer et à marcotter lui permettent d'installer de maintenir sa monodominance. Nous avons de plus développé un jeu de marqueurs microsatellites et constitué une banque de gènes au cours d'une importante campagne de terrain. Nous proposons une nouvelle classification de la monodominance afin de mieux appréhender les processus écologiques qui lui sont associés, et y replaçons le S. longifolia. Enfin, nous mettons en perspective nos résultats avec la gestion des peuplements naturels. / Large expanses of forest dominated by a single tree species, i.e. monodominant forests, occur through the tropics and remain an long-standing conundrum to most ecologists. In French Guiana, we described a new monodominant tree species: Spirotropis longifolia (DC) Baill. (Leguminosae-Papilionoideae), and studied, for the first time, its autecology. We reported a strong monodominant feature (up to 70 % of stems >10 cm in d.b.h.) and a marked difference between floristic composition of the dominated and adjacent, highly diverse, stands. The monodominance of S. longifolia was supported neither by peculiar soil conditions nor by a lack of competitors but may be owned to its astonishing self-coppicing and layering abilities which induce a sharp spatial structure. We also developed a set of microsatellite markers and conduct a large field survey to built up a gene data base. Facing the variety of ecological processes involved in monodominance, we identified a lack of conceptual framework, set a new classification of monodominance and positioned S. longifolia within. We finally discuss these results in the frame of the forest management.

No description available.