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.

Kinetic equations for difussion in the presence of entropic barriers

We use the mesoscopic nonequilibrium thermodynamics theory to derive the general kinetic equation of a system in the presence of potential barriers. The result is applied to a description of the evolution of systems whose dynamics is influenced by entropic barriers. We analyze in detail the case of diffusion in a domain of irregular geometry in which the presence of the boundaries induces an entropy barrier when approaching the exact dynamics by a coarsening of the description. The corresponding kinetic equation, named the Fick-Jacobs equation, is obtained, and its validity is generalized through the formulation of a scaling law for the diffusion coefficient which depends on the shape of the boundaries. The method we propose can be useful to analyze the dynamics of systems at the nanoscale where the presence of entropy barriers is a common feature.

An international university network for online teaching of pathology in Frenchspeaking countries

Introduction: Building online courses is a highly time consuming task for teachers of a single university. Universities working alone create high-quality courses but often cannot cover all pathological fields. Moreover this often leads to duplication of contents among universities, representing a big waste of teacher time and energy. We initiated in 2011 a French university network for building mutualized online teaching pathology cases, and this network has been extended in 2012 to Quebec and Switzerland. Method: Twenty French universities (see & for details), University Laval in Quebec and University of Lausanne in Switzerland are associated to this project. One e-learning Moodle platform (http://moodle.sorbonne-paris-cite.fr/) contains texts with URL pointing toward virtual slides that are decentralized in several universities. Each university has the responsibility of its own slide scanning, slide storage and online display with virtual slide viewers. The Moodle website is hosted by PRES Sorbonne Paris Cité, and financial supports for hardware have been obtained from UNF3S (http://www.unf3s.org/) and from PRES Sorbonne Paris Cité. Financial support for international fellowships has been obtained from CFQCU (http://www.cfqcu.org/). Results: The Moodle interface has been explained to pathology teachers using web-based conferences with screen sharing. The teachers added then contents such as clinical cases, selfevaluations and other media organized in several sections by student levels and pathological fields. Contents can be used as online learning or online preparation of subsequent courses in classrooms. In autumn 2013, one resident from Quebec spent 6 weeks in France and Switzerland and created original contents in inflammatory skin pathology. These contents are currently being validated by senior teachers and will be opened to pathology residents in spring 2014. All contents of the website can be accessed for free. Most contents just require anonymous connection but some specific fields, especially those containing pictures obtained from patients who agreed for a teaching use only, require personal identification of the students. Also, students have to register to access Moodle tests. All contents are written in French but one case has been translated into English to illustrate this communication (http://moodle.sorbonne-pariscite.fr/mod/page/view.php?id=261) (use "login as a guest"). The Moodle test module allows many types of shared questions, making it easy to create personalized tests. Contents that are opened to students have been validated by an editorial committee composed of colleagues from the participating institutions. Conclusions: Future developments include other international fellowships, the next one being scheduled for one French resident from May to October 2014 in Quebec, with a study program centered on lung and breast pathology. It must be kept in mind that these e-learning programs highly depend on teachers' time, not only at these early steps but also later to update the contents. We believe that funding resident fellowships for developing online pathological teaching contents is a win-win situation, highly beneficial for the resident who will improve his knowledge and way of thinking, highly beneficial for the teachers who will less worry about access rights or image formats, and finally highly beneficial for the students who will get courses fully adapted to their practice.

The promotion and assessment of generic skills from interdisciplinary teaching teams

The paper explains a teaching project financed by the University of Barcelona (UB). It focuses on ageneric skill of the University of Barcelona, which is defined as "the learning capability andresponsibility”, and in which analytical and synthesis skills are included. It follows a multidisciplinaryapproach including teachers of Mathematics, World Economics and Economic History. All of us sharethe same students during the first and the second course of the grade in Economics at the Faculty ofEconomics and Business. The project has been developed in three stages. The first one has beendone during the first semester of the course 2012/13, being applied to first year students on thesubjects of Mathematics and Economic History. The second phase is being to be done during thesecond semester only on the Economic History subject. A third stage is going to be done next course2013/14 to second year students on the subject of World Economics. Each different teaching teamhas developed specific materials and assessment tools for each one of the subjects included in theproject. The project emphasizes two teaching dimensions: the elaboration of teaching materials topromote the acquisition of generic skills from an interdisciplinary point of view, and the design ofspecific tools to assess such skills. The first results of the first phase of the project shows cleardeficiencies in the analytical skill regarding to first year students.

The promotion and assessment of generic skills from interdisciplinary teaching teams

The paper explains a teaching project financed by the University of Barcelona (UB). It focuses on ageneric skill of the University of Barcelona, which is defined as "the learning capability andresponsibility”, and in which analytical and synthesis skills are included. It follows a multidisciplinaryapproach including teachers of Mathematics, World Economics and Economic History. All of us sharethe same students during the first and the second course of the grade in Economics at the Faculty ofEconomics and Business. The project has been developed in three stages. The first one has beendone during the first semester of the course 2012/13, being applied to first year students on thesubjects of Mathematics and Economic History. The second phase is being to be done during thesecond semester only on the Economic History subject. A third stage is going to be done next course2013/14 to second year students on the subject of World Economics. Each different teaching teamhas developed specific materials and assessment tools for each one of the subjects included in theproject. The project emphasizes two teaching dimensions: the elaboration of teaching materials topromote the acquisition of generic skills from an interdisciplinary point of view, and the design ofspecific tools to assess such skills. The first results of the first phase of the project shows cleardeficiencies in the analytical skill regarding to first year students.

The promotion and assessment of generic skills from interdisciplinary teaching teams

The paper explains a teaching project financed by the University of Barcelona (UB). It focuses on ageneric skill of the University of Barcelona, which is defined as "the learning capability andresponsibility”, and in which analytical and synthesis skills are included. It follows a multidisciplinaryapproach including teachers of Mathematics, World Economics and Economic History. All of us sharethe same students during the first and the second course of the grade in Economics at the Faculty ofEconomics and Business. The project has been developed in three stages. The first one has beendone during the first semester of the course 2012/13, being applied to first year students on thesubjects of Mathematics and Economic History. The second phase is being to be done during thesecond semester only on the Economic History subject. A third stage is going to be done next course2013/14 to second year students on the subject of World Economics. Each different teaching teamhas developed specific materials and assessment tools for each one of the subjects included in theproject. The project emphasizes two teaching dimensions: the elaboration of teaching materials topromote the acquisition of generic skills from an interdisciplinary point of view, and the design ofspecific tools to assess such skills. The first results of the first phase of the project shows cleardeficiencies in the analytical skill regarding to first year students.

Psoriasis triggered by toll-like receptor 7 agonist imiquimod in the presence of dermal plasmacytoid dendritic cell precursors.

BACKGROUND: It has been proposed that the innate immune system plays a central role in driving the autoimmune T-cell cascade leading to psoriasis; however, there is no direct evidence for this. OBSERVATIONS: We observed aggravation and spreading of a psoriatic plaque when treated topically with the toll-like receptor (TLR) 7 agonist imiquimod. The exacerbation of psoriasis was accompanied by a massive induction of lesional type I interferon activity, detected by MxA expression after imiquimod therapy. Since imiquimod induces large amounts of type I interferon production from TLR7-expressing plasmacytoid dendritic cell precursors (PDCs), the natural interferon-producing cells of the peripheral blood, we asked whether PDCs are present in psoriatic skin. We identified high numbers of PDCs in psoriatic skin lesions (up to 16% of the total dermal infiltrate) based on their coexpression of BDCA2 and CD123. By contrast, PDCs were present at very low levels in atopic dermatitis and not detected in normal human skin. CONCLUSIONS: This study shows that psoriasis can be driven by the innate immune system through TLR ligation. Furthermore, our finding that large numbers of PDCs infiltrate psoriatic skin suggests a role of lesional PDCs as type I interferon-producing targets for the TLR7 agonist imiquimod.

Online teaching of inflammatory skin pathology by a French-speaking International University Network.

INTRODUCTION: Developments in technology, web-based teaching and whole slide imaging have broadened the teaching horizon in anatomic pathology. Creating online learning material including many types of media such as radiologic images, whole slides, videos, clinical and macroscopic photographs, is now accessible to most universities. Unfortunately, a major limiting factor to maintain and update the learning material is the amount of resources needed. In this perspective, a French-national university network was initiated in 2011 to build joint online teaching modules consisting of clinical cases and tests. The network has since expanded internationally to Québec, Switzerland and Ivory Coast. METHOD: One of the first steps of the project was to build a learning module on inflammatory skin pathology for interns and residents in pathology and dermatology. A pathology resident from Québec spent 6 weeks in France and Switzerland to develop the contents and build the module on an e-learning Moodle platform under the supervision of two dermatopathologists. The learning module contains text, interactive clinical cases, tests with feedback, virtual slides, images and clinical photographs. For that module, the virtual slides are decentralized in 2 universities (Bordeaux and Paris 7). Each university is responsible of its own slide scanning, image storage and online display with virtual slide viewers. RESULTS: The module on inflammatory skin pathology includes more than 50 web pages with French original content, tests and clinical cases, links to over 45 virtual images and more than 50 microscopic and clinical photographs. The whole learning module is being revised by four dermatopathologists and two senior pathologists. It will be accessible to interns and residents in the spring of 2014. The experience and knowledge gained from that work will be transferred to the next international resident whose work will be aimed at creating lung and breast pathology learning modules. CONCLUSION: The challenges of sustaining a project of this scope are numerous. The technical aspect of whole-slide imaging and storage needs to be developed by each university or group. The content needs to be regularly updated and its accuracy reviewed by experts in each individual domain. The learning modules also need to be promoted within the academic community to ensure maximal benefit for trainees. A collateral benefit of the project was the establishment of international partnerships between French-speaking universities and pathologists with the common goal of promoting pathology education through the use of multi-media technology including whole slide imaging.

Correlations between Vocal Input and Visual Response Apparently Enhance Presence in a Virtual Environment

This work investigates novel alternative means of interaction in a virtual environment (VE).We analyze whether humans can remap established body functions to learn to interact with digital information in an environment that is cross-sensory by nature and uses vocal utterances in order to influence (abstract) virtual objects. We thus establish a correlation among learning, control of the interface, and the perceived sense of presence in the VE. The application enables intuitive interaction by mapping actions (the prosodic aspects of the human voice) to a certain response (i.e., visualization). A series of single-user and multiuser studies shows that users can gain control of the intuitive interface and learn to adapt to new and previously unseen tasks in VEs. Despite the abstract nature of the presented environment, presence scores were generally very high.

Drug utilisation in a neonatal intensive care unit of a swiss university hospital

Introduction Medication errors in hospitalsmay occur at any step of the medication process including prescription, transcription, preparation and administration, and may originate with any of the actors involved. Neonatal intensive care units (NICU) take care of extremely frail patients in whom errors could have dramatic consequences. Our objective was to assess the frequency and nature of medication errors in the NICU of a university hospital in order to propose measures for improvement.Materials & Methods The design was that of an observational prospective study over 4 consecutivemonths. All patients receiving C 3drugs were included. For each patient, observations during the different stages were compiled in a computer formulary and compared with the litterature. Setting: The 11-bed NICU of our university hospital.Main outcome measures:(a) Frequency and nature of medication errors in prescription,transcription, preparation and administration.(b) Drugs affected by errors.Results 83 patients were included. 505 prescriptions and transcriptions, 447 preparations and 464 administrations were analyzed. 220 medications errors were observed: 102 (46.4%) at prescription, 25 (11.4%) at transcription, 19 (8.6%) at preparation and 73 (33.2%) at administration. Uncomplete/ambiguous orders (24; 23.5%) were the most common errors observed at prescription, followed by wrong name (21; 20.6%), wrong dose (17; 16.7%) and omission (15; 14.7%). Wrong time (33; 45.2%) and wrong administration technique (31; 42.5%) were the most important medication errors during administration. According to the ATC classification, systemic antibacterials (53; 24.1%) were the most implicated, followed by perfusion solutions (40; 18.2%), respiratory system products (30; 13.6%), and mineral supplements and antithrombotic agents (20; 9.1%).Discussions, Conclusion Proposed recommendations: ? Better teaching of neonatal prescription to medical interns;? Improved prescription form to avoid omissions and ambiguities;? Development of a neonatal drug formulary, including prescription,preparation and administration modalities to reduce errors at different stages;? Presence of a clinical pharmacist in the NICU.Disclosure of Interest None Declared

The JNK inhibitor XG-102 protects from ischemic damage with delayed intravenous administration also in the presence of recombinant tissue plasminogen activator.

BACKGROUND: XG-102 (formerly D-JNKI1), a TAT-coupled dextrogyre peptide which selectively inhibits the c-Jun N-terminal kinase, is a powerful neuroprotectant in mouse models of middle cerebral artery occlusion (MCAo) with delayed intracerebroventricular injection. We aimed to determine whether this neuroprotection could also be achieved by intravenous injection of XG-102, which is a more feasible approach for future use in stroke patients. We also tested the compatibility of the compound with recombinant tissue plasminogen activator (rtPA), commonly used for intravenous thrombolysis and known to enhance excitotoxicity. METHODS: Male ICR-CD1 mice were subjected to a 30-min-suture MCAo. XG-102 was injected intravenously in a single dose, 6 h after ischemia. Hippocampal slice cultures were subjected to oxygen (5%) and glucose (1 mM) deprivation for 30 min. rtPA was added after ischemia and before XG-102 administration, both in vitro and in vivo. RESULTS: The lowest intravenous dose achieving neuroprotection was 0.0003 mg/kg, which reduced the infarct volume after 48 h from 62 +/- 19 mm(3) (n = 18) for the vehicle-treated group to 18 +/- 9 mm(3) (n = 5, p < 0.01). The behavioral outcome was also significantly improved at two doses. Addition of rtPA after ischemia enhanced the ischemic damage both in vitro and in vivo, but XG-102 was still able to induce a significant neuroprotection. CONCLUSIONS: A single intravenous administration of XG-102 several hours after ischemia induces a powerful neuroprotection. XG-102 protects from ischemic damage in the presence of rtPA. The feasibility of systemic administration of this promising compound and its compatibility with rtPA are important steps for its development as a drug candidate in ischemic stroke.

Increased Presence of FOXP3+ Regulatory T Cells in Inflamed Muscle of Patients with Active Juvenile Dermatomyositis Compared to Peripheral Blood.

Juvenile dermatomyositis (JDM) is an immune-mediated inflammatory disease affecting the microvasculature of skin and muscle. CD4+CD25+FOXP3+ regulatory T cells (Tregs) are key regulators of immune homeostasis. A role for Tregs in JDM pathogenesis has not yet been established. Here, we explored Treg presence and function in peripheral blood and muscle of JDM patients. We analyzed number, phenotype and function of Tregs in blood from JDM patients by flow cytometry and in vitro suppression assays, in comparison to healthy controls and disease controls (Duchenne's Muscular Dystrophy). Presence of Tregs in muscle was analyzed by immunohistochemistry. Overall, Treg percentages in peripheral blood of JDM patients were similar compared to both control groups. Muscle biopsies of new onset JDM patients showed increased infiltration of numbers of T cells compared to Duchenne's muscular dystrophy. Both in JDM and Duchenne's muscular dystrophy the proportion of FOXP3+ T cells in muscles were increased compared to JDM peripheral blood. Interestingly, JDM is not a self-remitting disease, suggesting that the high proportion of Tregs in inflamed muscle do not suppress inflammation. In line with this, peripheral blood Tregs of active JDM patients were less capable of suppressing effector T cell activation in vitro, compared to Tregs of JDM in clinical remission. These data show a functional impairment of Tregs in a proportion of patients with active disease, and suggest a regulatory role for Tregs in JDM inflammation.

How we experience immersive virtual environments: the concept of presence and its measurement

This paper reviews the concept of presence in immersive virtual environments, the sense of being there signalled by people acting and responding realistically to virtual situations and events. We argue that presence is a unique phenomenon that must be distinguished from the degree of engagement, involvement in the portrayed environment. We argue that there are three necessary conditions for presence: the (a) consistent low latency sensorimotor loop between sensory data and proprioception; (b) statistical plausibility: images must be statistically plausible in relation to the probability distribution of images over natural scenes. A constraint on this plausibility is the level of immersion;(c) behaviour-response correlations: Presence may be enhanced and maintained over time by appropriate correlations between the state and behaviour of participants and responses within the environment, correlations that show appropriate responses to the activity of the participants. We conclude with a discussion of methods for assessing whether presence occurs, and in particular recommend the approach of comparison with ground truth and give some examples of this.

Faculty-librarians collaboration in an e-learning experience : resources management and training in Open University of Catalonia (UOC)

The traditional model of learning based on knowledge transfer doesn't promote the acquisition of information-related competencies and development of autonomous learning. More needs to be done to embrace learner-centred approaches, based on constructivism, collaboration and co-operation. This new learning paradigm is aligned with the European Higher Education Area (EHEA) requirements. In this sense, a learning experience based in faculty' librarian collaboration was seen as the best option for promoting student engagement and also a way to increase information-related competences in Open University of Catalonia (UOC) academic context. This case study outlines the benefits of teacher-librarian collaboration in terms of pedagogy innovation, resources management and introduction of open educational resources (OER) in virtual classrooms, Information literacy (IL) training and use of 2.0 tools in teaching. Our faculty-librarian's collaboration aims to provide an example of technology-enhanced learning and demonstrate how working together improves the quality and relevance of educational resources in UOC's virtual classrooms. Under this new approach, while teachers change their role from instructors to facilitators of the learning process and extend their reach to students, libraries acquire an important presence in the academic learning communities.