Machine learning for geospatial data : algorithms, software tools and case studies

Résumé Cette thèse est consacrée à l'analyse, la modélisation et la visualisation de données environnementales à référence spatiale à l'aide d'algorithmes d'apprentissage automatique (Machine Learning). L'apprentissage automatique peut être considéré au sens large comme une sous-catégorie de l'intelligence artificielle qui concerne particulièrement le développement de techniques et d'algorithmes permettant à une machine d'apprendre à partir de données. Dans cette thèse, les algorithmes d'apprentissage automatique sont adaptés pour être appliqués à des données environnementales et à la prédiction spatiale. Pourquoi l'apprentissage automatique ? Parce que la majorité des algorithmes d'apprentissage automatiques sont universels, adaptatifs, non-linéaires, robustes et efficaces pour la modélisation. Ils peuvent résoudre des problèmes de classification, de régression et de modélisation de densité de probabilités dans des espaces à haute dimension, composés de variables informatives spatialisées (« géo-features ») en plus des coordonnées géographiques. De plus, ils sont idéaux pour être implémentés en tant qu'outils d'aide à la décision pour des questions environnementales allant de la reconnaissance de pattern à la modélisation et la prédiction en passant par la cartographie automatique. Leur efficacité est comparable au modèles géostatistiques dans l'espace des coordonnées géographiques, mais ils sont indispensables pour des données à hautes dimensions incluant des géo-features. Les algorithmes d'apprentissage automatique les plus importants et les plus populaires sont présentés théoriquement et implémentés sous forme de logiciels pour les sciences environnementales. Les principaux algorithmes décrits sont le Perceptron multicouches (MultiLayer Perceptron, MLP) - l'algorithme le plus connu dans l'intelligence artificielle, le réseau de neurones de régression généralisée (General Regression Neural Networks, GRNN), le réseau de neurones probabiliste (Probabilistic Neural Networks, PNN), les cartes auto-organisées (SelfOrganized Maps, SOM), les modèles à mixture Gaussiennes (Gaussian Mixture Models, GMM), les réseaux à fonctions de base radiales (Radial Basis Functions Networks, RBF) et les réseaux à mixture de densité (Mixture Density Networks, MDN). Cette gamme d'algorithmes permet de couvrir des tâches variées telle que la classification, la régression ou l'estimation de densité de probabilité. L'analyse exploratoire des données (Exploratory Data Analysis, EDA) est le premier pas de toute analyse de données. Dans cette thèse les concepts d'analyse exploratoire de données spatiales (Exploratory Spatial Data Analysis, ESDA) sont traités selon l'approche traditionnelle de la géostatistique avec la variographie expérimentale et selon les principes de l'apprentissage automatique. La variographie expérimentale, qui étudie les relations entre pairs de points, est un outil de base pour l'analyse géostatistique de corrélations spatiales anisotropiques qui permet de détecter la présence de patterns spatiaux descriptible par une statistique. L'approche de l'apprentissage automatique pour l'ESDA est présentée à travers l'application de la méthode des k plus proches voisins qui est très simple et possède d'excellentes qualités d'interprétation et de visualisation. Une part importante de la thèse traite de sujets d'actualité comme la cartographie automatique de données spatiales. Le réseau de neurones de régression généralisée est proposé pour résoudre cette tâche efficacement. Les performances du GRNN sont démontrées par des données de Comparaison d'Interpolation Spatiale (SIC) de 2004 pour lesquelles le GRNN bat significativement toutes les autres méthodes, particulièrement lors de situations d'urgence. La thèse est composée de quatre chapitres : théorie, applications, outils logiciels et des exemples guidés. Une partie importante du travail consiste en une collection de logiciels : Machine Learning Office. Cette collection de logiciels a été développée durant les 15 dernières années et a été utilisée pour l'enseignement de nombreux cours, dont des workshops internationaux en Chine, France, Italie, Irlande et Suisse ainsi que dans des projets de recherche fondamentaux et appliqués. Les cas d'études considérés couvrent un vaste spectre de problèmes géoenvironnementaux réels à basse et haute dimensionnalité, tels que la pollution de l'air, du sol et de l'eau par des produits radioactifs et des métaux lourds, la classification de types de sols et d'unités hydrogéologiques, la cartographie des incertitudes pour l'aide à la décision et l'estimation de risques naturels (glissements de terrain, avalanches). Des outils complémentaires pour l'analyse exploratoire des données et la visualisation ont également été développés en prenant soin de créer une interface conviviale et facile à l'utilisation. Machine Learning for geospatial data: algorithms, software tools and case studies Abstract The thesis is devoted to the analysis, modeling and visualisation of spatial environmental data using machine learning algorithms. In a broad sense machine learning can be considered as a subfield of artificial intelligence. It mainly concerns with the development of techniques and algorithms that allow computers to learn from data. In this thesis machine learning algorithms are adapted to learn from spatial environmental data and to make spatial predictions. Why machine learning? In few words most of machine learning algorithms are universal, adaptive, nonlinear, robust and efficient modeling tools. They can find solutions for the classification, regression, and probability density modeling problems in high-dimensional geo-feature spaces, composed of geographical space and additional relevant spatially referenced features. They are well-suited to be implemented as predictive engines in decision support systems, for the purposes of environmental data mining including pattern recognition, modeling and predictions as well as automatic data mapping. They have competitive efficiency to the geostatistical models in low dimensional geographical spaces but are indispensable in high-dimensional geo-feature spaces. The most important and popular machine learning algorithms and models interesting for geo- and environmental sciences are presented in details: from theoretical description of the concepts to the software implementation. The main algorithms and models considered are the following: multi-layer perceptron (a workhorse of machine learning), general regression neural networks, probabilistic neural networks, self-organising (Kohonen) maps, Gaussian mixture models, radial basis functions networks, mixture density networks. This set of models covers machine learning tasks such as classification, regression, and density estimation. Exploratory data analysis (EDA) is initial and very important part of data analysis. In this thesis the concepts of exploratory spatial data analysis (ESDA) is considered using both traditional geostatistical approach such as_experimental variography and machine learning. Experimental variography is a basic tool for geostatistical analysis of anisotropic spatial correlations which helps to understand the presence of spatial patterns, at least described by two-point statistics. A machine learning approach for ESDA is presented by applying the k-nearest neighbors (k-NN) method which is simple and has very good interpretation and visualization properties. Important part of the thesis deals with a hot topic of nowadays, namely, an automatic mapping of geospatial data. General regression neural networks (GRNN) is proposed as efficient model to solve this task. Performance of the GRNN model is demonstrated on Spatial Interpolation Comparison (SIC) 2004 data where GRNN model significantly outperformed all other approaches, especially in case of emergency conditions. The thesis consists of four chapters and has the following structure: theory, applications, software tools, and how-to-do-it examples. An important part of the work is a collection of software tools - Machine Learning Office. Machine Learning Office tools were developed during last 15 years and was used both for many teaching courses, including international workshops in China, France, Italy, Ireland, Switzerland and for realizing fundamental and applied research projects. Case studies considered cover wide spectrum of the real-life low and high-dimensional geo- and environmental problems, such as air, soil and water pollution by radionuclides and heavy metals, soil types and hydro-geological units classification, decision-oriented mapping with uncertainties, natural hazards (landslides, avalanches) assessments and susceptibility mapping. Complementary tools useful for the exploratory data analysis and visualisation were developed as well. The software is user friendly and easy to use.

Environmental data mining and modeling based on machine learning algorithms and geostatistics

The paper presents some contemporary approaches to spatial environmental data analysis. The main topics are concentrated on the decision-oriented problems of environmental spatial data mining and modeling: valorization and representativity of data with the help of exploratory data analysis, spatial predictions, probabilistic and risk mapping, development and application of conditional stochastic simulation models. The innovative part of the paper presents integrated/hybrid model-machine learning (ML) residuals sequential simulations-MLRSS. The models are based on multilayer perceptron and support vector regression ML algorithms used for modeling long-range spatial trends and sequential simulations of the residuals. NIL algorithms deliver non-linear solution for the spatial non-stationary problems, which are difficult for geostatistical approach. Geostatistical tools (variography) are used to characterize performance of ML algorithms, by analyzing quality and quantity of the spatially structured information extracted from data with ML algorithms. Sequential simulations provide efficient assessment of uncertainty and spatial variability. Case study from the Chernobyl fallouts illustrates the performance of the proposed model. It is shown that probability mapping, provided by the combination of ML data driven and geostatistical model based approaches, can be efficiently used in decision-making process. (C) 2003 Elsevier Ltd. All rights reserved.

Multi-scale systems analysis of plant development : beyond cellular signaling and tissue morphodynamics

Les plantes sont essentielles pour les sociétés humaines. Notre alimentation quotidienne, les matériaux de constructions et les sources énergétiques dérivent de la biomasse végétale. En revanche, la compréhension des multiples aspects développementaux des plantes est encore peu exploitée et représente un sujet de recherche majeur pour la science. L'émergence des technologies à haut débit pour le séquençage de génome à grande échelle ou l'imagerie de haute résolution permet à présent de produire des quantités énormes d'information. L'analyse informatique est une façon d'intégrer ces données et de réduire la complexité apparente vers une échelle d'abstraction appropriée, dont la finalité est de fournir des perspectives de recherches ciblées. Ceci représente la raison première de cette thèse. En d'autres termes, nous appliquons des méthodes descriptives et prédictives combinées à des simulations numériques afin d'apporter des solutions originales à des problèmes relatifs à la morphogénèse à l'échelle de la cellule et de l'organe. Nous nous sommes fixés parmi les objectifs principaux de cette thèse d'élucider de quelle manière l'interaction croisée des phytohormones auxine et brassinosteroïdes (BRs) détermine la croissance de la cellule dans la racine du méristème apical d'Arabidopsis thaliana, l'organisme modèle de référence pour les études moléculaires en plantes. Pour reconstruire le réseau de signalement cellulaire, nous avons extrait de la littérature les informations pertinentes concernant les relations entre les protéines impliquées dans la transduction des signaux hormonaux. Le réseau a ensuite été modélisé en utilisant un formalisme logique et qualitatif pour pallier l'absence de données quantitatives. Tout d'abord, Les résultats ont permis de confirmer que l'auxine et les BRs agissent en synergie pour contrôler la croissance de la cellule, puis, d'expliquer des observations phénotypiques paradoxales et au final, de mettre à jour une interaction clef entre deux protéines dans la maintenance du méristème de la racine. Une étude ultérieure chez la plante modèle Brachypodium dystachion (Brachypo- dium) a révélé l'ajustement du réseau d'interaction croisée entre auxine et éthylène par rapport à Arabidopsis. Chez ce dernier, interférer avec la biosynthèse de l'auxine mène à la formation d'une racine courte. Néanmoins, nous avons isolé chez Brachypodium un mutant hypomorphique dans la biosynthèse de l'auxine qui affiche une racine plus longue. Nous avons alors conduit une analyse morphométrique qui a confirmé que des cellules plus anisotropique (plus fines et longues) sont à l'origine de ce phénotype racinaire. Des analyses plus approfondies ont démontré que la différence phénotypique entre Brachypodium et Arabidopsis s'explique par une inversion de la fonction régulatrice dans la relation entre le réseau de signalisation par l'éthylène et la biosynthèse de l'auxine. L'analyse morphométrique utilisée dans l'étude précédente exploite le pipeline de traitement d'image de notre méthode d'histologie quantitative. Pendant la croissance secondaire, la symétrie bilatérale de l'hypocotyle est remplacée par une symétrie radiale et une organisation concentrique des tissus constitutifs. Ces tissus sont initialement composés d'une douzaine de cellules mais peuvent aisément atteindre des dizaines de milliers dans les derniers stades du développement. Cette échelle dépasse largement le seuil d'investigation par les moyens dits 'traditionnels' comme l'imagerie directe de tissus en profondeur. L'étude de ce système pendant cette phase de développement ne peut se faire qu'en réalisant des coupes fines de l'organe, ce qui empêche une compréhension des phénomènes cellulaires dynamiques sous-jacents. Nous y avons remédié en proposant une stratégie originale nommée, histologie quantitative. De fait, nous avons extrait l'information contenue dans des images de très haute résolution de sections transverses d'hypocotyles en utilisant un pipeline d'analyse et de segmentation d'image à grande échelle. Nous l'avons ensuite combiné avec un algorithme de reconnaissance automatique des cellules. Cet outil nous a permis de réaliser une description quantitative de la progression de la croissance secondaire révélant des schémas développementales non-apparents avec une inspection visuelle classique. La formation de pôle de phloèmes en structure répétée et espacée entre eux d'une longueur constante illustre les bénéfices de notre approche. Par ailleurs, l'exploitation approfondie de ces résultats a montré un changement de croissance anisotropique des cellules du cambium et du phloème qui semble en phase avec l'expansion du xylème. Combinant des outils génétiques et de la modélisation biomécanique, nous avons démontré que seule la croissance plus rapide des tissus internes peut produire une réorientation de l'axe de croissance anisotropique des tissus périphériques. Cette prédiction a été confirmée par le calcul du ratio des taux de croissance du xylème et du phloème au cours de développement secondaire ; des ratios élevés sont effectivement observés et concomitant à l'établissement progressif et tangentiel du cambium. Ces résultats suggèrent un mécanisme d'auto-organisation établi par un gradient de division méristématique qui génèrent une distribution de contraintes mécaniques. Ceci réoriente la croissance anisotropique des tissus périphériques pour supporter la croissance secondaire. - Plants are essential for human society, because our daily food, construction materials and sustainable energy are derived from plant biomass. Yet, despite this importance, the multiple developmental aspects of plants are still poorly understood and represent a major challenge for science. With the emergence of high throughput devices for genome sequencing and high-resolution imaging, data has never been so easy to collect, generating huge amounts of information. Computational analysis is one way to integrate those data and to decrease the apparent complexity towards an appropriate scale of abstraction with the aim to eventually provide new answers and direct further research perspectives. This is the motivation behind this thesis work, i.e. the application of descriptive and predictive analytics combined with computational modeling to answer problems that revolve around morphogenesis at the subcellular and organ scale. One of the goals of this thesis is to elucidate how the auxin-brassinosteroid phytohormone interaction determines the cell growth in the root apical meristem of Arabidopsis thaliana (Arabidopsis), the plant model of reference for molecular studies. The pertinent information about signaling protein relationships was obtained through the literature to reconstruct the entire hormonal crosstalk. Due to a lack of quantitative information, we employed a qualitative modeling formalism. This work permitted to confirm the synergistic effect of the hormonal crosstalk on cell elongation, to explain some of our paradoxical mutant phenotypes and to predict a novel interaction between the BREVIS RADIX (BRX) protein and the transcription factor MONOPTEROS (MP),which turned out to be critical for the maintenance of the root meristem. On the same subcellular scale, another study in the monocot model Brachypodium dystachion (Brachypodium) revealed an alternative wiring of auxin-ethylene crosstalk as compared to Arabidopsis. In the latter, increasing interference with auxin biosynthesis results in progressively shorter roots. By contrast, a hypomorphic Brachypodium mutant isolated in this study in an enzyme of the auxin biosynthesis pathway displayed a dramatically longer seminal root. Our morphometric analysis confirmed that more anisotropic cells (thinner and longer) are principally responsible for the mutant root phenotype. Further characterization pointed towards an inverted regulatory logic in the relation between ethylene signaling and auxin biosynthesis in Brachypodium as compared to Arabidopsis, which explains the phenotypic discrepancy. Finally, the morphometric analysis of hypocotyl secondary growth that we applied in this study was performed with the image-processing pipeline of our quantitative histology method. During its secondary growth, the hypocotyl reorganizes its primary bilateral symmetry to a radial symmetry of highly specialized tissues comprising several thousand cells, starting with a few dozens. However, such a scale only permits observations in thin cross-sections, severely hampering a comprehensive analysis of the morphodynamics involved. Our quantitative histology strategy overcomes this limitation. We acquired hypocotyl cross-sections from tiled high-resolution images and extracted their information content using custom high-throughput image processing and segmentation. Coupled with an automated cell type recognition algorithm, it allows precise quantitative characterization of vascular development and reveals developmental patterns that were not evident from visual inspection, for example the steady interspace distance of the phloem poles. Further analyses indicated a change in growth anisotropy of cambial and phloem cells, which appeared in phase with the expansion of xylem. Combining genetic tools and computational modeling, we showed that the reorientation of growth anisotropy axis of peripheral tissue layers only occurs when the growth rate of central tissue is higher than the peripheral one. This was confirmed by the calculation of the ratio of the growth rate xylem to phloem throughout secondary growth. High ratios are indeed observed and concomitant with the homogenization of cambium anisotropy. These results suggest a self-organization mechanism, promoted by a gradient of division in the cambium that generates a pattern of mechanical constraints. This, in turn, reorients the growth anisotropy of peripheral tissues to sustain the secondary growth.

Gastrointestinal transit of extruded or pelletized diets in pacu fed distinct inclusion levels of lipid and carbohydrate

The objective of this work was to evaluate the effect of pelletized or extruded diets, with different levels of carbohydrate and lipid, on the gastrointestinal transit time (GITT) and its modulation in pacu (Piaractus mesopotamicus). One hundred and eighty pacu juveniles were fed with eight isonitrogenous diets containing two carbohydrate levels (40 and 50%) and two lipid levels (4 and 8%). Four diets were pelletized and four were extruded. Carbohydrate and lipid experimental levels caused no changes to the bolus transit time. However, the bolus permanence time was related to diet processing. Fish fed pelletized diets exhibited the highest gastrointestinal transit time. Regression analysis of bolus behavior for pelletized and extruded diets with 4% lipid depicted different fits. GITT regression analysis of fish fed 8% lipid was fitted to a cubic equation and displayed adjustments of food permanence, with enhanced utilization of the diets, either with extruded or pelletized diets. GITT of fish fed extruded diets with 4% lipid was adjusted to a linear equation. The GITT of pacu depends on the diet processing and is affected by dietary levels of lipid and carbohydrate.

MOCVD growth andproperties of GaN and AlGaN layers

In this work GaN and AlGaN layers were grown by metal-organic chemical vapor deposition (MOCVD) on sapphire substrates. The research was carried out at Micro and Nanoscience Laboratory of Helsinki University of Technology. The objective of this thesis is the study of MOCVD technique for the growth of GaN and AlGaN films and optimization of growth parameters in purpose to improve crystal quality of the films. The widely used two-step and the new multistep methods have been used for GaN, AlGaN MOCVD growth on c-plane sapphire. Properties of the GaN and AlGaN layers were studied using in-situ reflectance monitoring during MOCVD growth, atomic force microscopy and x-ray diffraction. Compared to the two step method, the multistep method has produced even better qualities of the GaN and AlGaN layers and significant reduction of threading dislocation density.

The use of positional scanning synthetic combinatorial libraries (PS-SCL) to study T Lympohcyte specificity and degeneracy

Les cellules CD8? T cytolytiques (CTL) sont les principaux effecteurs du système immunitaire adaptatif contre les infections et les tumeurs. La récente identification d?antigènes tumoraux humains reconnus par des cellules T cytolytiques est la base pour le, développement des vaccins antigène spécifiques contre le cancer. Le nombre d?antigènes tumoraux reconnus par des CTL que puisse être utilisé comme cible pour la vaccination des patients atteints du cancer est encore limité. Une nouvelle technique, simple et rapide, vient d?être proposée pour l?identification d?antigènes reconnus par des CTL. Elle se base sur l?utilisation de librairies combinatoriales de peptides arrangées en un format de "scanning" ou balayage par position (PS-SCL). La première partie de cette étude a consisté à valider cette nouvelle technique par une analyse détaillée de la reconnaissance des PS-SCL par différents clones de CTL spécifiques pour des antigènes associés à la tumeur (TAA) connus ainsi que par des clones de spécificité inconnue. Les résultats de ces analyses révèlent que pour tous les clones, la plupart des acides aminés qui composent la séquence du peptide antigénique naturel ont été identifiés par l?utilisation des PS-SCL. Les résultats obtenus ont permis d?identifier des peptides analogues ayant une antigènicité augmentée par rapport au peptide naturel, ainsi que des peptides comportant de multiples modifications de séquence, mais présentant la même réactivité que le peptide naturel. La deuxième partie de cette étude a consisté à effectuer des analyses biométriques des résultats complexes générés par la PS-SCL. Cette approche a permis l?identification des séquences correspondant aux épitopes naturels à partir de bases de données de peptides publiques. Parmi des milliers de peptides, les séquences naturelles se trouvent comprises dans les 30 séquences ayant les scores potentiels de stimulation les plus élevés pour chaque TAA étudié. Mais plus important encore, l?utilisation des PS-SCL avec un clone réactif contre des cellules tumorales mais de spécificité inconnue nous a permis d?identifier I?epitope reconnu par ce clone. Les données présentées ici encouragent l?utilisation des PS-SCL pour l?identification et l?optimisation d?épitopes pour des CTL réactifs anti-tumoraux, ainsi que pour l?étude de la reconnaissance dégénérée d?antigènes par les CTL.<br/><br/>CD8+ cytolytic T lymphocytes (CTL) are the main effector cells of the adaptive immune system against infection and tumors. The recent identification of moleculariy defined human tumor Ags recognized by autologous CTL has opened new opportunities for the development of Ag-specific cancer vaccines. Despite extensive work, however, the number of CTL-defined tumor Ags that are suitable targets for the vaccination of cancer patients is still limited, especially because of the laborious and time consuming nature of the procedures currentiy used for their identification. The use of combinatorial peptide libraries in positionai scanning format (Positional Scanning Synthetic Combinatorial Libraries, PS-SCL)' has recently been proposed as an alternative approach for the identification of these epitopes. To validate this approach, we analyzed in detail the recognition of PS-SCL by tumor-reactive CTL clones specific for multiple well-defined tumor-associated Ags (TAA) as well as by tumor-reactive CTL clones of unknown specificity. The results of these analyses revealed that for all the TAA-specific clones studied most of the amino acids composing the native antigenic peptide sequences could be identified through the use of PS-SCL. Based on the data obtained from the screening of PS-SCL, we could design peptide analogs of increased antigenicity as well as cross-reactive analog peptides containing multiple amino acid substitutions. In addition, the resuits of PS-SCL-screening combined with a recently developed biometric data analysis (PS-SCL-based biometric database analysis) allowed the identification of the native peptides in public protein databases among the 30 most active sequences, and this was the case for all the TAA studied. More importantiy, the screening of PS- SCL with a tumor-reactive CTL clone of unknown specificity resulted in the identification of the actual epitope. Overall, these data encourage the use of PS-SCL not oniy for the identification and optimization of tumor-associated CTL epitopes, but also for the analysis of degeneracy in T lymphocyte receptor (TCR) recognition of tumor Ags.<br/><br/>Les cellules T CD8? cytolytiques font partie des globules blancs du sang et sont les principales responsables de la lutte contre les infections et les tumeurs. Les immunologistes cherchent depuis des années à identifier des molécules exprimées et présentées à la surface des tumeurs qui puissent être reconnues par des cellules T CD8? cytolytiques capables ensuite de tuer ces tumeurs de façon spécifique. Ce type de molécules représente la base pour le développement de vaccins contre le cancer puisqu?elles pourraient être injectées aux patients afin d?induire une réponse anti- tumorale. A présent, il y a très peu de molécules capables de stimuler le système immunitaire contre les tumeurs qui sont connues parce que les techniques développées à ce jour pour leur identification sont complexes et longues. Une nouvelle technique vient d?être proposée pour l?identification de ce type de molécules qui se base sur l?utilisation de librairies de peptides. Ces librairies représentent toutes les combinaisons possibles des composants de base des molécules recherchées. La première partie de cette étude a consisté à valider cette nouvelle technique en utilisant des cellules T CD8? cytolytiques capables de tuer des cellules tumorales en reconnaissant une molécule connue présente à leur surface. On a démontré que l?utilisation des librairies permet d?identifier la plupart des composants de base de la molécule reconnue par les cellules T CD8? cytolytiques utilisées. La deuxième partie de cette étude a consisté à effectuer une recherche des molécules potentiellement actives dans des protéines présentes dans des bases des données en utilisant un programme informatique qui permet de classer les molécules sur la base de leur activité biologique. Parmi des milliers de molécules de la base de données, celles reconnues par nos cellules T CD8? cytolytiques ont été trouvées parmi les plus actives. Plus intéressant encore, la combinaison de ces deux techniques nous a permis d?identifier la molécule reconnue par une population de cellules T CD8? cytolytiques ayant une activité anti-tumorale, mais pour laquelle on ne connaissait pas la spécificité. Nos résultats encouragent l?utilisation des librairies pour trouver et optimiser des molécules reconnues spécifiquement par des cellules T CD8? cytolytiques capables de tuer des tumeurs.

Local and Global Feature Extraction for Invariant Object Recognition

Perceiving the world visually is a basic act for humans, but for computers it is still an unsolved problem. The variability present innatural environments is an obstacle for effective computer vision. The goal of invariant object recognition is to recognise objects in a digital image despite variations in, for example, pose, lighting or occlusion. In this study, invariant object recognition is considered from the viewpoint of feature extraction. Thedifferences between local and global features are studied with emphasis on Hough transform and Gabor filtering based feature extraction. The methods are examined with respect to four capabilities: generality, invariance, stability, and efficiency. Invariant features are presented using both Hough transform and Gabor filtering. A modified Hough transform technique is also presented where the distortion tolerance is increased by incorporating local information. In addition, methods for decreasing the computational costs of the Hough transform employing parallel processing and local information are introduced.

Measuring irregularities and surface defects from printed patterns

Quality inspection and assurance is a veryimportant step when today's products are sold to markets. As products are produced in vast quantities, the interest to automate quality inspection tasks has increased correspondingly. Quality inspection tasks usuallyrequire the detection of deficiencies, defined as irregularities in this thesis. Objects containing regular patterns appear quite frequently on certain industries and science, e.g. half-tone raster patterns in the printing industry, crystal lattice structures in solid state physics and solder joints and components in the electronics industry. In this thesis, the problem of regular patterns and irregularities is described in analytical form and three different detection methods are proposed. All the methods are based on characteristics of Fourier transform to represent regular information compactly. Fourier transform enables the separation of regular and irregular parts of an image but the three methods presented are shown to differ in generality and computational complexity. Need to detect fine and sparse details is common in quality inspection tasks, e.g., locating smallfractures in components in the electronics industry or detecting tearing from paper samples in the printing industry. In this thesis, a general definition of such details is given by defining sufficient statistical properties in the histogram domain. The analytical definition allowsa quantitative comparison of methods designed for detail detection. Based on the definition, the utilisation of existing thresholding methodsis shown to be well motivated. Comparison of thresholding methods shows that minimum error thresholding outperforms other standard methods. The results are successfully applied to a paper printability and runnability inspection setup. Missing dots from a repeating raster pattern are detected from Heliotest strips and small surface defects from IGT picking papers.

The UlaG protein familly defines novel structural and functional motifs grafted on an ancient RNase fold

Background: Bacterial populations are highly successful at colonizing new habitats and adapting to changing environmental conditions, partly due to their capacity to evolve novel virulence and metabolic pathways in response to stress conditions and to shuffle them by horizontal gene transfer (HGT). A common theme in the evolution of new functions consists of gene duplication followed by functional divergence. UlaG, a unique manganese-dependent metallo-b-lactamase (MBL) enzyme involved in L-ascorbate metabolism by commensal and symbiotic enterobacteria, provides a model for the study of the emergence of new catalytic activities from the modification of an ancient fold. Furthermore, UlaG is the founding member of the so-called UlaG-like (UlaGL) protein family, a recently established and poorly characterized family comprising divalent (and perhaps trivalent)metal-binding MBLs that catalyze transformations on phosphorylated sugars and nucleotides. Results: Here we combined protein structure-guided and sequence-only molecular phylogenetic analyses to dissect the molecular evolution of UlaG and to study its phylogenomic distribution, its relatedness with present-day UlaGL protein sequences and functional conservation. Phylogenetic analyses indicate that UlaGL sequences are present in Bacteria and Archaea, with bona fide orthologs found mainly in mammalian and plant-associated Gramnegative and Gram-positive bacteria. The incongruence between the UlaGL tree and known species trees indicates exchange by HGT and suggests that the UlaGL-encoding genes provided a growth advantage under changing conditions. Our search for more distantly related protein sequences aided by structural homology has uncovered that UlaGL sequences have a common evolutionary origin with present-day RNA processing and metabolizing MBL enzymes widespread in Bacteria, Archaea, and Eukarya. This observation suggests an ancient origin for the UlaGL family within the broader trunk of the MBL superfamily by duplication, neofunctionalization and fixation. Conclusions: Our results suggest that the forerunner of UlaG was present as an RNA metabolizing enzyme in the last common ancestor, and that the modern descendants of that ancestral gene have a wide phylogenetic distribution and functional roles. We propose that the UlaGL family evolved new metabolic roles among bacterial and possibly archeal phyla in the setting of a close association with metazoans, such as in the mammalian gastrointestinal tract or in animal and plant pathogens, as well as in environmental settings. Accordingly, the major evolutionary forces shaping the UlaGL family include vertical inheritance and lineage-specific duplication and acquisition of novel metabolic functions, followed by HGT and numerous lineage-specific gene loss events.

Cooperation versus competition effects on information sharing and use in group decision making

Information processing in groups has long been seen as a cooperative process. In contrast with this assumption, group members were rarely found to behave cooperatively: They withhold unshared information and stick to initial incorrect decisions. In the present article, we examined how group members' cooperative and competitivemotives impact on group information processing and propose that information sharing and use in groups could be seen as strategic behavior. We reviewed the latest developments in the literature investigating different forms of strategic information processing and their underlying mechanisms. This review suggests that explicit cooperative goals are needed for effective group decision-making.

The empirical assessment of agency accountability: A regime approach and an application to the German Bundesnetzagentur

Regulation has in many cases been delegated to independent agencies, which has led to the question of how democratic accountability of these agencies is ensured. There are few empirical approaches to agency accountability. We offer such an approach, resting upon three propositions. First, we scrutinize agency accountability both de jure (accountability is ensured by formal rights of accountability 'fora' to receive information and impose consequences) and de facto (the capability of fora to use these rights depends on resources and decision costs that affect the credibility of their sanctioning capacity). Second, accountability must be evaluated separately at political, operational and managerial levels. And third, at each level accountability is enacted by a system of several (partially) interdependent fora, forming together an accountability regime. The proposed framework is applied to the case of the German Bundesnetzagentur's accountability regime, which shows its suitability for empirical purposes. Regulatory agencies are often considered as independent, yet accountable. This article provides a realistic framework for the study of accountability 'regimes' in which they are embedded. It emphasizes the need to identify the various actors (accountability fora) to which agencies are formally accountable (parliamentary committees, auditing bodies, courts, and so on) and to consider possible relationships between them. It argues that formal accountability 'on paper', as defined in official documents, does not fully account for de facto accountability, which depends on the resources possessed by the fora (mainly information-processing and decision-making capacities) and the credibility of their sanctioning capacities. The article applies this framework to the German Bundesnetzagentur.

Electrostimulation mapping of comprehension of auditory and visual words.

In order to spare functional areas during the removal of brain tumours, electrical stimulation mapping was used in 90 patients (77 in the left hemisphere and 13 in the right; 2754 cortical sites tested). Language functions were studied with a special focus on comprehension of auditory and visual words and the semantic system. In addition to naming, patients were asked to perform pointing tasks from auditory and visual stimuli (using sets of 4 different images controlled for familiarity), and also auditory object (sound recognition) and Token test tasks. Ninety-two auditory comprehension interference sites were observed. We found that the process of auditory comprehension involved a few, fine-grained, sub-centimetre cortical territories. Early stages of speech comprehension seem to relate to two posterior regions in the left superior temporal gyrus. Downstream lexical-semantic speech processing and sound analysis involved 2 pathways, along the anterior part of the left superior temporal gyrus, and posteriorly around the supramarginal and middle temporal gyri. Electrostimulation experimentally dissociated perceptual consciousness attached to speech comprehension. The initial word discrimination process can be considered as an "automatic" stage, the attention feedback not being impaired by stimulation as would be the case at the lexical-semantic stage. Multimodal organization of the superior temporal gyrus was also detected since some neurones could be involved in comprehension of visual material and naming. These findings demonstrate a fine graded, sub-centimetre, cortical representation of speech comprehension processing mainly in the left superior temporal gyrus and are in line with those described in dual stream models of language comprehension processing.

Research networks and inventors’ mobility as drivers of innovation evidence from Europe

We investigate the importance of the labour mobility of inventors, as well as the scale, extent and density of their collaborative research networks, for regional innovation outcomes. To do so, we apply a knowledge production function framework at the regional level and include inventors’ networks and their labour mobility as regressors. Our empirical approach takes full account of spatial interactions by estimating a spatial lag model together, where necessary, with a spatial error model. In addition, standard errors are calculated using spatial heteroskedasticity and autocorrelation consistent estimators to ensure their robustness in the presence of spatial error autocorrelation and heteroskedasticity of unknown form. Our results point to the existence of a robust positive correlation between intraregional labour mobility and regional innovation, whilst the relationship with networks is less clear. However, networking across regions positively correlates with a region’s innovation intensity.

General Equilibrium Long-Run Determinants for Spanish FDI: A Spatial Panel Data Approach

While general equilibrium theories of trade stress the role of third-country effects, little work has been done in the empirical foreign direct investment (FDI) literature to test such spatial linkages. This paper aims to provide further insights into long-run determinants of Spanish FDI by considering not only bilateral but also spatially weighted third-country determinants. The few studies carried out so far have focused on FDI flows in a limited number of countries. However, Spanish FDI outflows have risen dramatically since 1995 and today account for a substantial part of global FDI. Therefore, we estimate recently developed Spatial Panel Data models by Maximum Likelihood (ML) procedures for Spanish outflows (1993-2004) to top-50 host countries. After controlling for unobservable effects, we find that spatial interdependence matters and provide evidence consistent with New Economic Geography (NEG) theories of agglomeration, mainly due to complex (vertical) FDI motivations. Spatial Error Models estimations also provide illuminating results regarding the transmission mechanism of shocks.

Fourier self-deconvolution in coherent anti-Stokes Raman scattering spectrum analysis

Coherent anti-Stokes Raman scattering (CARS) microscopy is rapidly developing into a unique microscopic tool in biophysics, biology and the material sciences. The nonlinear nature of CARS spectroscopy complicates the analysis of the received spectra. There were developed mathematical methods for signal processing and for calculations spectra. Fourier self-deconvolution is a special high pass FFT filter which synthetically narrows the effective trace bandwidth features. As Fourier self-deconvolution can effectively reduce the noise, which may be at a higher spatial frequency than the peaks, without losing peak resolution. The idea of the work is to experiment the possibility of using wavelet decomposition in spectroscopic for background and noise removal, and Fourier transformation for linenarrowing.