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.

Olfactory traces and spatial learning in rats.

We conducted an experiment to assess the use of olfactory traces for spatial orientation in an open environment in rats, Rattus norvegicus. We trained rats to locate a food source at a fixed location from different starting points, in the presence or absence of visual information. A single food source was hidden in an array of 19 petri dishes regularly arranged in an open-field arena. Rats were trained to locate the food source either in white light (with full access to distant visuospatial information) or in darkness (without any visual information). In both cases, the goal was in a fixed location relative to the spatial frame of reference. The results of this experiment revealed that the presence of noncontrolled olfactory traces coherent with the spatial frame of reference enables rats to locate a unique position as accurately in darkness as with full access to visuospatial information. We hypothesize that the olfactory traces complement the use of other orientation mechanisms, such as path integration or the reliance on visuospatial information. This experiment demonstrates that rats can rely on olfactory traces for accurate orientation, and raises questions about the establishment of such traces in the absence of any other orientation mechanism. Copyright 1998 The Association for the Study of Animal Behaviour.

Family relationships and learning experiences in anxious vs. non-anxious families: Are there differences ?

L'objectif principal de ce travail était d'explorer les relations parent-enfant et les processus d'apprentissage familiaux associés aux troubles anxieux. A cet effet, des familles ayant un membre anxieux (la mère ou l'enfant) ont été comparées avec des familles n'ayant aucun membre anxieux. Dans une première étude, l'observation de l'interaction mère-enfant, pendant une situation standardisée de jeu, a révélé que les mères présentant un trouble panique étaient plus susceptibles de se montrer verbalement contrôlantes, critiques et moins sensibles aux besoins de l'enfant, que les mères qui ne présentaient pas de trouble panique. Une deuxième étude a examiné les perceptions des différents membres de la famille quant aux relations au sein de la famille et a indiqué que, par comparaison aux adolescents non-anxieux, les adolescents anxieux étaient plus enclins à éprouver un sentiment d'autonomie individuelle diminué par rapport à leurs parents. Finalement, une troisième étude s'est intéressée à déterminer l'impact d'expériences d'apprentissage moins directes dans l'étiologie de l'anxiété. Les résultats ont indiqué que les mères présentant un trouble panique étaient plus enclines à s'engager dans des comportements qui maintiennent la panique et à impliquer leurs enfants dans ces comportements, que les mères ne présentant pas de trouble panique. En se basant sur des recherches antérieures qui ont établi une relation entre le contrôle parental, la perception de contrôle chez l'enfant et les troubles anxieux, le présent travail non seulement confirme ce lien mais propose également un modèle pour résumer l'état actuel des connaissances concernant les processus familiaux et le développement des troubles anxieux. Deux routes ont été suggérées par lesquelles l'anxiété pourrait être transmise de manière intergénérationnelle. Chacune de ces routes attribue un rôle important à la perception de contrôle chez l'enfant. L'idée est que lorsque les enfants présentent une prédisposition à interpréter le comportement de leurs parents comme hors de leur contrôle, ils seraient plus enclins à développer de l'anxiété. A ce titre, la perception du contrôle représenterait un tampon entre le comportement de contrôle/surprotection des parents et le trouble anxieux chez l'enfant. - The principal objective of the present work was to explore parent-child relationships and family learning processes associated with anxiety disorders. To this purpose, families with and without an anxious family member (mother or child) were compared. In a first study, observation of mother-child interaction, during a standard play situation, revealed that mothers with panic disorder were more likely to display verbal control and criticism, and less likely to display sensitivity toward their children than mothers without panic disorder. A second study examined family members' perceptions of family relationships and indicated that compared to non-anxious adolescents, anxious adolescents were more prone to experience a diminished sense of individual autonomy in relation to their parents. Finally a third study was interested in determining the effect of less direct learning experiences in the aetiology of anxiety. Results indicated that mothers with panic disorder were more likely to engage in panic-maintaining behaviour and to involve their children in this behaviour than mothers without panic disorder. Based on previous research showing a relationship between parental control, children's perception of control, and anxiety disorders, the present work not only further adds evidence to support this link but also proposes a model summarizing the current knowledge concerning family processes and the development of anxiety disorders. Two pathways have been suggested through which anxiety may be intergenerationally transmitted. Both pathways assign an important role to children's perception of control. The idea is that whenever children have a predisposition towards interpreting their parents' behaviour as beyond of their control, they may be more prone to develop anxiety. As such, perceived control may represent a buffer between parental overcontrolling/overprotective behaviours and childhood anxiety disorder.

Using active learning for monitoring networks design: The example of wind power plants sites evaluation

Locating new wind farms is of crucial importance for energy policies of the next decade. To select the new location, an accurate picture of the wind fields is necessary. However, characterizing wind fields is a difficult task, since the phenomenon is highly nonlinear and related to complex topographical features. In this paper, we propose both a nonparametric model to estimate wind speed at different time instants and a procedure to discover underrepresented topographic conditions, where new measuring stations could be added. Compared to space filling techniques, this last approach privileges optimization of the output space, thus locating new potential measuring sites through the uncertainty of the model itself.

Influence of local environmental olfactory cues on place learning in rats

The aim of the present study was to assess the influence of local environmental olfactory cues on place learning in rats. We developed a new experimental design allowing the comparison of the use of local olfactory and visual cues in spatial and discrimination learning. We compared the effect of both types of cues on the discrimination of a single food source in an open-field arena. The goal was either in a fixed or in a variable location, and could be indicated by local olfactory and/or visual cues. The local cues enhanced the discrimination of the goal dish, whether it was in a fixed or in a variable location. However, we did not observe any overshadowing of the spatial information by the local olfactory or visual cue. Rats relied primarily on distant visuospatial information to locate the goal, neglecting local information when it was in conflict with the spatial information.

A behavioral study of alpha-1b adrenergic receptor knockout mice: increased reaction to novelty and selectively reduced learning capacities.

Knockout mice lacking the alpha-1b adrenergic receptor were tested in behavioral experiments. Reaction to novelty was first assessed in a simple test in which the time taken by the knockout mice and their littermate controls to enter a second compartment was compared. Then the mice were tested in an open field to which unknown objects were subsequently added. Special novelty was introduced by moving one of the familiar objects to another location in the open field. Spatial behavior and memory were further studied in a homing board test, and in the water maze. The alpha-1b knockout mice showed an enhanced reactivity to new situations. They were faster to enter the new environment, covered longer paths in the open field, and spent more time exploring the new objects. They reacted like controls to modification inducing spatial novelty. In the homing board test, both the knockout mice and the control mice seemed to use a combination of distant visual and proximal olfactory cues, showing place preference only if the two types of cues were redundant. In the water maze the alpha-1b knockout mice were unable to learn the task, which was confirmed in a probe trial without platform. They were perfectly able, however, to escape in a visible platform procedure. These results confirm previous findings showing that the noradrenergic pathway is important for the modulation of behaviors such as reaction to novelty and exploration, and suggest that this is mediated, at least partly, through the alpha-1b adrenergic receptors. The lack of alpha-1b adrenergic receptors in spatial orientation does not seem important in cue-rich tasks but may interfere with orientation in situations providing distant cues only.

Learning-induced plasticity in human audition: Objects, time, and space.

The human auditory system is comprised of specialized but interacting anatomic and functional pathways encoding object, spatial, and temporal information. We review how learning-induced plasticity manifests along these pathways and to what extent there are common mechanisms subserving such plasticity. A first series of experiments establishes a temporal hierarchy along which sounds of objects are discriminated along basic to fine-grained categorical boundaries and learned representations. A widespread network of temporal and (pre)frontal brain regions contributes to object discrimination via recursive processing. Learning-induced plasticity typically manifested as repetition suppression within a common set of brain regions. A second series considered how the temporal sequence of sound sources is represented. We show that lateralized responsiveness during the initial encoding phase of pairs of auditory spatial stimuli is critical for their accurate ordered perception. Finally, we consider how spatial representations are formed and modified through training-induced learning. A population-based model of spatial processing is supported wherein temporal and parietal structures interact in the encoding of relative and absolute spatial information over the initial ∼300ms post-stimulus onset. Collectively, these data provide insights into the functional organization of human audition and open directions for new developments in targeted diagnostic and neurorehabilitation strategies.

A Survey of Active Learning Algorithms for Supervised Remote Sensing Image Classification

Defining an efficient training set is one of the most delicate phases for the success of remote sensing image classification routines. The complexity of the problem, the limited temporal and financial resources, as well as the high intraclass variance can make an algorithm fail if it is trained with a suboptimal dataset. Active learning aims at building efficient training sets by iteratively improving the model performance through sampling. A user-defined heuristic ranks the unlabeled pixels according to a function of the uncertainty of their class membership and then the user is asked to provide labels for the most uncertain pixels. This paper reviews and tests the main families of active learning algorithms: committee, large margin, and posterior probability-based. For each of them, the most recent advances in the remote sensing community are discussed and some heuristics are detailed and tested. Several challenging remote sensing scenarios are considered, including very high spatial resolution and hyperspectral image classification. Finally, guidelines for choosing the good architecture are provided for new and/or unexperienced user.