Les techniques avancées pour l’approximation de rang faible des matrices sont des outils de réduction de dimension fondamentaux pour un grand nombre de domaines du calcul scientifique. Les approches hiérarchiques comme les matrices H2, en particulier la méthode multipôle rapide (FMM), bénéficient de la structure de rang faible par bloc de certaines matrices pour réduire le coût de calcul de problèmes d’interactions à n-corps en O(n) opérations au lieu de O(n2). Afin de mieux traiter des noyaux d’interaction complexes de plusieurs natures, des formulations FMM dites ”kernel-independent” ont récemment vu le jour, telles que les FMM basées sur l’interpolation polynomiale. Cependant elles deviennent très coûteuses pour les noyaux tensoriels à fortes dimensions, c’est pourquoi nous avons développé une nouvelle formulation FMM efficace basée sur l’interpolation polynomiale, appelée Uniform FMM. Cette méthode a été implémentée dans la bibliothèque parallèle ScalFMM et repose sur une grille d’interpolation régulière et la transformée de Fourier rapide (FFT). Ses performances et sa précision ont été comparées à celles de la FMM par interpolation de Chebyshev. Des simulations numériques sur des cas tests artificiels ont montré que la perte de précision induite par le schéma d’interpolation était largement compensées par le gain de performance apporté par la FFT. Dans un premier temps, nous avons étendu les FMM basées sur grille de Chebyshev et sur grille régulière au calcul des champs élastiques isotropes mis en jeu dans des simulations de Dynamique des Dislocations (DD). Dans un second temps, nous avons utilisé notre nouvelle FMM pour accélérer une factorisation SVD de rang r par projection aléatoire et ainsi permettre de générer efficacement des champs Gaussiens aléatoires sur de grandes grilles hétérogènes. Pour finir, nous avons développé un algorithme de réduction de dimension basé sur la projection aléatoire dense afin d’étudier de nouvelles façons de caractériser la biodiversité, à savoir d’un point de vue géométrique. / Advanced techniques for the low-rank approximation of matrices are crucial dimension reduction tools in many domains of modern scientific computing. Hierarchical approaches like H2-matrices, in particular the Fast Multipole Method (FMM), benefit from the block low-rank structure of certain matrices to reduce the cost of computing n-body problems to O(n) operations instead of O(n2). In order to better deal with kernels of various kinds, kernel independent FMM formulations have recently arisen such as polynomial interpolation based FMM. However, they are hardly tractable to high dimensional tensorial kernels, therefore we designed a new highly efficient interpolation based FMM, called the Uniform FMM, and implemented it in the parallel library ScalFMM. The method relies on an equispaced interpolation grid and the Fast Fourier Transform (FFT). Performance and accuracy were compared with the Chebyshev interpolation based FMM. Numerical experiments on artificial benchmarks showed that the loss of accuracy induced by the interpolation scheme was largely compensated by the FFT optimization. First of all, we extended both interpolation based FMM to the computation of the isotropic elastic fields involved in Dislocation Dynamics (DD) simulations. Second of all, we used our new FMM algorithm to accelerate a rank-r Randomized SVD and thus efficiently generate multivariate Gaussian random variables on large heterogeneous grids in O(n) operations. Finally, we designed a new efficient dimensionality reduction algorithm based on dense random projection in order to investigate new ways of characterizing the biodiversity, namely from a geometric point of view.

The aim of this thesis is comparison of speakers' understanding of meaning of names of body parts head, eye, ear, nose, mouth and heart in Czech and Japanese language. To achieve this goal a list of idioms consisting of mentioned body parts in both languages is created. Speakers sort the idioms into arbitrary categories based on their understaing of meaning of name of body part. Resulting data is processed by multidimensional scaling and evaluated afterwards. The first bigger chapter is dedicated to introduction of idiomatics and phraseology in Europe and Japan in order to compare understanding of idioms and their classification. The second bigger chapter introduces a base for approaches to idioms, following by focus on somatic idioms. Two different approaches are introduced. The method of introspection is represented by a theory of coneptual profiles by I. Vaňková and by an idealized cognitive model of act by Arizono T. The corpus approach is represented by F. Čermák, Hashimoto Ch. et al. and by Lin W. The introduction of practical part follows. The first part is dedicated to the method of multidimensional scaling. After that the survey realization and its evaluation is described. The overall evaluation of results in terms of the aim of this thesis is placed in conclusion. Keywords: Japanese...

Besides the social survey data, texts have been an important source of sociological data since the beginning of the development of sociological methodology. Text analysis methods contain two main branches of development: Bernard Berelson's content analysis and Hans-Georg Gadamer's hermeneutic analysis. Both these methodological branches have been influenced by the development of information technologies in the last twenty years. The thesis presented here deals with one of the methods of computer text analysis (CATA), which stands on the border between these two methodological streams, a method of analyzing words' collocations in texts. The thesis presents the method in the context of other methods of text analysis, and mentions sources of inspiration for further development of these methods - corpus linguistics and text mining. The second part discusses the different steps of words' collocation analysis: building a text corpus, dictionary compilation, calculation of data matrix and visualisation of words' distances using multidimensional scaling (MDS). The method is also applied to a specific data, two text corpora compiled from transcripts of biographical interviews with actors of Czechoslovak normalization - with dissidents and Communist functionaries. Quality of the models is assessed, depending...

In order to understand Mio-Pliocene floristic characteristics of the southern Appalachian Mountains, 47 palynological samples from six different testing-pits across the Gray Fossil Site (GFS) were analyzed. The site exhibits a low pollen yield resulting from basic pH levels, drought, and fire events occurring during deposition. The palynofloral assemblage has a low to moderate diversity, and it is largely dominated by Quercus-Carya-Pinus (~90% of the palynoflora). The reported presence of Pterocarya grains supports a Late Neogene age for the lacustrine sediments. Comparison with modern pollen-based floras suggests that: (1) the Mio-Pliocene Oak-Hickory-Pine association varied in structure between a woodland to woodland/savanna, depending on the intensity and frequency of drought, fire events and herbivory, (2) pits show different structure of the co-dominant genera, which indicate alternating composition of the vegetation, and (3) in term of modern vegetation, the GFS flora corresponds well with the Mesophytic Forest region.

State and Federal agencies have created sanctuaries and speed zones to help reduce manatee mortality while incorporating the recreational and commercial resource needs of these same habitats for humans. Specific habitat resources are considered necessary to increase manatee survivorship. We have only recently begun to address how manatees use some of these resources based on physiological or reproductive strategies. In this study, I quantified patterns of habitat use during seasonal and diel periods for different sex and reproductive manatee classes using data from a radio-telemetry study conducted by the Florida Fish and Wildlife Conservation Commission during 1991-1996. I used five environmental geographic data layers: bathymetry, distance to seagrass, distance to shoreline, distance to warm water refuge sites, and distance to fresh water sources, to discriminate seasonal and diel habitat use patterns for different manatee classes: males (M), females with calves (FWC), and females without calves (FNC). Mean occupancy values were calculated for environmental variable locations and seasonal, diel, and manatee class differences were tested using a Multi-Response Permutation Procedure (MRPP). Nonmetric Multidimensional Scaling (NMS) was used to visualize the ordination patterns of the manatee classes and to assess importance of correlated environmental variables. Significant differences in habitat use were noted between summer and winter based on distances to warm water, seagrass, and fresh water sources but similar habitat use patterns were exhibited within summer diel periods among manatee classes. All manatee classes appeared to have used a higher proportion of locations closer in proximity to seagrass at night than day in winter indicating a disproportionate difference in feeding bouts between diel periods. These differences may be attributed to adjusting feeding strategies to reduce thermoregulatory costs or to decrease human interactions. Differences in patterns were exhibited for the winter diel periods specifically for the FWC manatee classes during winter days. FWC had a higher proportion of locations within the warm water refuges during the day indicating a possible trade off situation between food consumption and thermal exposure. This study demonstrates coarse and fine scale patterns of variation in habitat use for manatees both seasonally and daily within winter. It also suggests that during winter months, manatees were not just utilizing their habitat but they appeared to have preferences and selection for certain habitat types. Recovery of a species is greatly enhanced when patterns of habitat use within the species' environment has been clearly defined. Understanding more specifically what types of habitats manatees choose might allow management to adjust strategies for protection of key habitats while encouraging further recovery of this species.

The vegetation responses to seven silvicultural treatments one growing season after harvesting were examined on seven sites in the southern Appalachian mountains of Virginia and West Virginia. Treatments included: 1) control, 2) understory control by herbicide, 3) group selection, 4) high-leave shelterwood, 5) low-leave shelterwood, 6) leave tree, and 7) clearcut. The effects of harvesting were compared between treatments and between pre-harvest and post-harvest samplings. Species richness, percent cover, and local species extinctions were calculated for sample plots ranging in size from 1m2 to 2 ha. Vegetation richness and cover increased with increasing harvest intensity. Local species extinctions were similar in the control and disturbed treatments. Additional analyses were performed using the control, high-leave shelterwood, and clearcut on five of the seven sites to determine the relationships between soil, litter, and other environmental characteristics and vegetation in the herbaceous layer (<1 m in height). Multivariate analysis techniques were used to analyze average differences in species abundance between pre-harvest and post-harvest and to relate post-harvest vegetation to microsite characteristics. Regional-scale differences in site location were more important in explaining the presence of a species than were environmental characteristics. Within a region, species primarily were distributed along a light/litter weight gradient and secondarily along a soil properties and nutrient gradient. / Master of Science

L’action humaine dans une séquence vidéo peut être considérée comme un volume spatio- temporel induit par la concaténation de silhouettes dans le temps. Nous présentons une approche spatio-temporelle pour la reconnaissance d’actions humaines qui exploite des caractéristiques globales générées par la technique de réduction de dimensionnalité MDS et un découpage en sous-blocs afin de modéliser la dynamique des actions. L’objectif est de fournir une méthode à la fois simple, peu dispendieuse et robuste permettant la reconnaissance d’actions simples. Le procédé est rapide, ne nécessite aucun alignement de vidéo, et est applicable à de nombreux scénarios. En outre, nous démontrons la robustesse de notre méthode face aux occultations partielles, aux déformations de formes, aux changements d’échelle et d’angles de vue, aux irrégularités dans l’exécution d’une action, et à une faible résolution. / Human action in a video sequence can be seen as a space-time volume induced by the concatenation of silhouettes in time. We present a space-time approach for human action recognition, which exploits global characteristics generated by the technique of dimensionality reduction MDS and a cube division into sub-blocks to model the dynamics of the actions. The objective is to provide a method that is simple, inexpensive and robust allowing simple action recognition. The process is fast, does not require video alignment, and is applicable in many scenarios. Moreover, we demonstrate the robustness of our method to partial occlusion, deformation of shapes, significant changes in scale and viewpoint, irregularities in the performance of an action, and low-quality video.

Biological processes are so complicated that to understand the mechanisms underlying the functioning of biomolecules it is inevitable to study them from various perspectives and with a wide range of tools. Understanding the function at the molecular level obviously requires the knowledge of the three dimensional structure of the biomolecules. Experimentally this can be obtained by techniques such as X‐ray crystallography and NMR studies. Computational biology has also played an important role in elucidating the structure function relationship in biomolecules. Computationally one can obtain the temporal as well as ensemble behavior of biomolecules at atomic level under conditions that are experimentally not accessible. Molecular dynamics(MD) study is a technique that can be used to obtain information of the dynamic behavior of the biomolecules. Dynamics of large systems like proteins can be investigated by classical force fields. However, the changes at the level of covalent bond involve the reorganization of electron density distribution which can be addressed only at Quantum mechanical level. In the present thesis, some of the biological systems have been characterized both at the classical and quantum mechanical level. The systems investigated by MD simulations and the insights brought from these studies are presented in Chapters 3 and 4. The unusual bonds such as pyrophosphate linkage in ATP and short strong hydrogen bonds in proteins, investigated through high level quantum chemical methods, are presented in Chapters 5, 6 and 7. Part of this thesis is aimed to address some important issues related to the dynamics of Tryptophanyl tRNA synthetase (TrpRS) which belongs to classic of aminoacyl‐tRNA synthetases (aaRS). aaRSs are extremely important class of enzymes involved in the translation of genetic code. These enzymes catalyze the aminoacylation of tRNAs to relate the cognate amino acids to the anticodon trinucleotide sequences. aaRSs are modular enzymes with distinct domains on which extensive kinetic and mutational experiments as well as structural analyses have been carried out, highlighting the role of inter‐domain communication (Alexander and Schimmel, 2001). The overall architecture of tRNA synthetases consists of primarily two domains. The active site domain is responsible for the activation of an amino acid with ATP in synthesizing an enzyme‐bound aminoacyl‐adenylate, and transfer of the aminoacyl‐adenylate intermediate to the 3’end of tRNA. The second domain is responsible for selection and binding of the cognate tRNA. aaRSs are allosteric proteins in which the binding of tRNA at the anticodon domain influences the activity at the catalytic region. These two binding sites are separated by a large distance. One of the aims of this thesis is to characterize such long distance communication (allosteric communication) at atomic level in Tryptophanyl tRNA synthetase. This is achieved by generating ensembles of conformations by MD simulations and analyzing the trajectories by novel graph theoretic approach. Graph and network based approaches are well established in the field of protein structure analysis for analyzing protein structure, stability and function (Kannan and Vishveshwara, 1999; Brinda and Vishveshwara, 2005). The parameters such as clusters, hubs and shortest paths provide valuable information on the structure and dynamics of the proteins. In this thesis, network parameters are used for the analysis of molecular dynamics MD) simulation data, to represent the global dynamic behavior of protein in a more elegant way. MD simulations are performed on some available (and modeled) structures of TrpRS bound to a variety of ligands, and the protein structure networks( PSN) of non‐covalent interactions are characterized in dynamical equilibrium. The ligand induced conformational changes are investigated through structure networks. These networks are used to understand the mode of communication between the anticodon domain and the active site. The interface dynamics is crucial for the function of TrpRS (since it is a functional dimer) and it is investigated through interface clusters. The matter embodied in the thesis is presented as 9 chapters. Chapter 1 lays the suitable background and foundation for the study, surveying relevant literature from different fields .Chapter 2 describes in detail the various materials, methods and techniques employed in the different analyses and studies presented in this thesis. A brief description of well‐known methods of molecular dynamics simulations, essential dynamics calculations, cross correlation maps, conformational clustering etc.is presented. The methods for constructing protein structure graphs and networks, developed in our lab, are described in detail. The use of network parameters for the analysis of MD simulation data to address the problem of communication between the two distal sites is also presented. Some descriptions of the ab initio quantum mechanical methods, which are used to investigate the unusual bonds in biomolecules, are also presented in this chapter. Chapter 3 is devoted in discussing the results from several normal as well as high temperature MD simulations of ligand‐free and ligand bound Bacillus stearothermophilus Tryptophanyl‐tRNA synthetase (bsTrpRS). The essential modes of the protein in the presence of different ligands are captured by essential dynamics calculations. Different conformations of the protein associated with the catalysis process of TrpRS, as captured through experiments, are discussed in the context of conformational sampling. High temperature simulations are carried out to explore the larger conformational space. Chapter 4 is focused on the results obtained from the MD simulation of human Tryptophanyl‐tRNA synthetase (hTrpRS). The structure of human TrpRS bound to the activated ligand (TrpAMP) and the cognate tRNA(tRNATRP) is modeled since no structure in the presence of both TrpAMP and tRNATRP is available. MD simulations on these modeled as well as other complexes of hTrpRS are performed to capture the dynamical process of ligand induced conformational changes (Hansiaetal., communicated). Both the local and the global changes in the protein conformation from the protein structure network (PSN) of MD snapshots are analyzed. Several important information such as the ligand induced correlation between different residues of the protein, asymmetric binding of the ligands to the two subunits of the protein, and the path of communication between the anticodon region and the aminoacylation site are obtained. Also, the role of the dimmer interface, from a dynamic perspective, is obtained for the first time. The interface dynamics which stabilize different quaternary structures of lectins (with high sequence and structure similarity) were investigated in a collaborative work (Hansiaetal.,2007). The lectin peanut agglutinin (PNA) is a tetramer with three different types of interfaces. The interface dynamics of this protein in the presence and in the absence of metal ions was investigated and the paper reporting the results from this study is included as appendix in this thesis. Chapter 5 deals with high level ab initio quantum chemical calculations on tri‐ and diphosphate fragments of adenosine triphosphate (ATP). Pyrophosphate prototypes such as methyl triphosphate and methyl diphosphate molecules in their different protonation states have been investigated at high levels of calculations (Hansiaetal., 2006a). The optimized geometries, the thermochemistry of the hydrolysis and the molecular orbitals contributing to the high energy of these compounds have been analyzed. These investigations provide insights into the‘‘highenergy’’character of ATP molecule. Further, the dependence of vibrational frequencies on the number of phosphate groups and the charged states has also been presented. These results aid in the interpretation of spectra obtained by experiments on complexes containing pyrophosphate prototypes. Hydrogen bonding is fundamental in understanding the structure and properties of molecules of biological interest including proteins. A recent analysis carried out in our lab showed that a significant number of short hydrogen bonds (SHB) are present in proteins (Rajagopal and Vishveshwara, 2005). Chapters 6 and 7 elucidate the results obtained from ab initio quantum chemical calculations on some of these SHBs to get aquantitative estimation of their geometry and strength. In chapter 6, asystematic analysis of the geometries and the energetics of possible SHB systems, which are frequently encountered in proteins, are presented at different levels of theory (HF,DFTandMP2). It is found that the SHBs involving both charged residues in the proteins are intrinsic in nature. However, two neutral residues form a SHB in the protein crystal structures either due to geometric constraints or due to the environment of these residues. This analysis enables one to distinguish SHBs which are formed because of geometric constraints from those which are formed because of the inherent property of the chemical groups involved in the hydrogen bonding. These results are useful in refining protein structures determined by crystallographic or NMR methods. In addition, sulfur atom of methionine and cysteinein proteins also participate in SHBs, which are not so well characterized. Chapter 7 presents the similar analysis carried out on short hydrogen bonds in proteins involving sulfur atom. A detailed analysis of SHBs of sulfur containing groups in a data set of proteins has been carried out. Some of the residue pairs from this analysis were considered for ab initio calculations. However, the optimization of these examples resulted in breaking of the hydrogen bonds involving sulfur atoms and formation of new hydrogen bonds with oxygen and/or nitrogen atoms. Hence model systems, which mimic the real examples, were designed to carry out ab initio studies and to investigate the short hydrogen bonds involving sulfur atoms. Another study on the protein‐water interaction, which does not fall under the realm of the main objective of the thesis, is discussed in Chapter 8. Protein–water interaction is crucial for accomplishing many biological functions of proteins. In the recent past, natural probe tryptophan, located at the protein surfaces, has been extensively investigated using femtosecond spectroscopy experiments to understand salvation dynamics (Peonetal.,2002). In this chapter a method is described to follow up the molecular events of the protein–water interactions in detail. Tryptophan–water interaction in the protein Monellin is investigated in order to get the atomic level insights into the hydration dynamics, by carrying out MD simulations on Monellin (Hansiaetal.,2006b). The results are compared with those obtained from femtosecond resolved fluorescence spectroscopy. The time constants of the survival correlation function match well with the reported experimental values.This validates the procedure, adapted here for Monellin, to investigate the hydration dynamics in general. The last chapter (Chapter9) summarizes the results obtained from various studies and discusses the future directions. First part of this thesis aims to present the analysis by carrying out MD simulations on monomeric and dimeric TrpRS protein in order to understand the two steps of the aminoacylation reaction: activation of the aminoacid Trp in the first step and the transfer of the activated amino acid in the next step. In the second part, quantitative estimation of the geometry and the strength of pyrophosphate bond and short hydrogen bonds in proteins are reported in detail by subjecting the systems to high levels of quantum mechanical calculations(QM). The use of ab initio QM/MM calculations by combining the quantum mechanics(QM) with the molecular mechanics(MM) in order to study the enzymatic reactions is discussed as the future

Grasses are widely used in revegetation to control erosion, build soil and maintain habitat. In northeast British Columbia, non-native grass species are commonly seeded to reclaim industrially disturbed sites. Widespread concern about degradation of biodiversity and key ecological processes has led to increasing value placed on native species and management practices leading to a more resilient landscape. I undertook this study to fill the restoration knowledge gap relating to native grasses in northeast BC. I did an extensive inventory of grasses on 217 sites in 2007, 2008 and 2009. Functional traits were measured in the field and in a greenhouse growth experiment. I found ninety-nine grass species occuring in the region, 70% of which are native. The number, proportion and extent of non-native grasses are increasing and four of these – Poa pratensis, Festuca rubra, Bromus inermis, and Phleum pratense represented almost a quarter of all occurrences. Several native species were common throughout the region: Calamagrostis canadensis, Leymus innovatus, Elymus trachycaulus, Poa palustris and Agrostis scabra. Other native species, including Festuca altaica, Koeleria macrantha, Pascopyrum smithii, and Schizachne purpurascens, strongly favoured intact habitats. Elevation, soil moisture regime, proportion of bare ground, and land use were significant factors related to local grass species composition and abundance. Agrostis scabra, Alopecurus aequalis, Beckmannia syzigachne, Bromus ciliatus, Cinna latifolia, Deschampsia cespitosa, Elymus alaskanus, Elymus trachycaulus, Festuca saximontana and Hordeum jubatum grew commonly on severely damaged well sites. Field measurements for Specific Leaf Area (SLA) and Leaf Dry Matter Content (LDMC) of 11 species showed an inverse correlation. Bromus ciliatus, Bromus pumpellianus, and Elymus trachycaulus had high SLA/low LDMC linked to rapid growth, whereas Festuca altaica, Deschampsia cespitosa, and Calamagrostis stricta had low SLA/high LDMC linked to slow growth and persistence. In the greenhouse experiment, Poa palustris, Cinna latifolia and Bromus ciliatus produced the most overall biomass and Pascopyrum smithii and Poa palustris produced the greatest aboveground biomass. Calamagrostis stricta, Poa palustris, Elymus glaucus, Leymus innovatus and Pascopyrum smithii exhibited clonal growth. Beckmannia syzigachne, Bromus ciliatus Cinna latifolia produced viable seed during the 135-day experiment. Considering all attributes five native species, Calamagrostis canadensis, Elymus trachycaulus, Poa palustris, Leymus innovatus, and Agrostis scabra are recommended for general restoration use in northeast B.C. Other native species show promise when matched to particular site conditions, including Alopecurus aequalis, Arctagrostis latifolia, Beckmannia syzigachne, Bromus ciliatus, Calamagrostis stricta, Cinna latifolia, Deschampsia cespitosa, Elymus glaucus, Festuca saximontana, Glyceria striata, Hordeum jubatum, Koeleria macrantha, Pascopyrum smithii, Poa alpina, Schizachne purpurascens and Trisetum spicatum. This information will be valuable to land managers interested in moving beyond reclamation to ecological restoration of sites disturbed by oil and gas development. Developing practices that are environmentally sound and socially acceptable requires ongoing botanical inventory. Plant traits may be useful in matching species to site conditions and restoration goals. Policy recommendations include phasing in of requirements to use native seed while restricting the use of agronomic species, promoting natural colonization, and supporting a native seed industry.

No description available.