On Recurrent and Deep Neural Networks

L'apprentissage profond est un domaine de recherche en forte croissance en apprentissage automatique qui est parvenu à des résultats impressionnants dans différentes tâches allant de la classification d'images à la parole, en passant par la modélisation du langage. Les réseaux de neurones récurrents, une sous-classe d'architecture profonde, s'avèrent particulièrement prometteurs. Les réseaux récurrents peuvent capter la structure temporelle dans les données. Ils ont potentiellement la capacité d'apprendre des corrélations entre des événements éloignés dans le temps et d'emmagasiner indéfiniment des informations dans leur mémoire interne. Dans ce travail, nous tentons d'abord de comprendre pourquoi la profondeur est utile. Similairement à d'autres travaux de la littérature, nos résultats démontrent que les modèles profonds peuvent être plus efficaces pour représenter certaines familles de fonctions comparativement aux modèles peu profonds. Contrairement à ces travaux, nous effectuons notre analyse théorique sur des réseaux profonds acycliques munis de fonctions d'activation linéaires par parties, puisque ce type de modèle est actuellement l'état de l'art dans différentes tâches de classification. La deuxième partie de cette thèse porte sur le processus d'apprentissage. Nous analysons quelques techniques d'optimisation proposées récemment, telles l'optimisation Hessian free, la descente de gradient naturel et la descente des sous-espaces de Krylov. Nous proposons le cadre théorique des méthodes à région de confiance généralisées et nous montrons que plusieurs de ces algorithmes développés récemment peuvent être vus dans cette perspective. Nous argumentons que certains membres de cette famille d'approches peuvent être mieux adaptés que d'autres à l'optimisation non convexe. La dernière partie de ce document se concentre sur les réseaux de neurones récurrents. Nous étudions d'abord le concept de mémoire et tentons de répondre aux questions suivantes: Les réseaux récurrents peuvent-ils démontrer une mémoire sans limite? Ce comportement peut-il être appris? Nous montrons que cela est possible si des indices sont fournis durant l'apprentissage. Ensuite, nous explorons deux problèmes spécifiques à l'entraînement des réseaux récurrents, à savoir la dissipation et l'explosion du gradient. Notre analyse se termine par une solution au problème d'explosion du gradient qui implique de borner la norme du gradient. Nous proposons également un terme de régularisation conçu spécifiquement pour réduire le problème de dissipation du gradient. Sur un ensemble de données synthétique, nous montrons empiriquement que ces mécanismes peuvent permettre aux réseaux récurrents d'apprendre de façon autonome à mémoriser des informations pour une période de temps indéfinie. Finalement, nous explorons la notion de profondeur dans les réseaux de neurones récurrents. Comparativement aux réseaux acycliques, la définition de profondeur dans les réseaux récurrents est souvent ambiguë. Nous proposons différentes façons d'ajouter de la profondeur dans les réseaux récurrents et nous évaluons empiriquement ces propositions.

A Diagnostic and Modeling Study on the Onset of Summer Monsoon over South Asia

This study focuses on the onset of southwest monsoon over Kerala. India Meteorological Department (IMD) has been using a semi-objective method to define monsoon onset. The main objectives of the study are to understand the monsoon onset processes, to simulate monsoon onset in a GCM using as input the atmospheric conditions and Sea Surface Temperature, 10 days earlier to the onset, to develop a method for medium range prediction of the date of onset of southwest monsoon over Kerala and to examine the possibility of objectively defining the date of Monsoon Onset over Kerala (MOK). It gives a broad description of regional monsoon systems and monsoon onsets over Asia and Australia. Asian monsoon includes two separate subsystems, Indain monsoon and East Asian monsoon. It is seen from this study that the duration of the different phases of the onset process are dependent on the period of ISO. Based on the study of the monsoon onset process, modeling studies can be done for better understanding of the ocean-atmosphere interaction especially those associated with the warm pool in the Bay of Bengal and the Arabian Sea.

Seasonal Studies on Primary Production and Associated Environmental Parameters in the Indian EEZ

The thesis describes the importance of Indian EEZ, definition and the various factors affecting primary production, general account of phytoplankton and its importance in marine ecosystem etc. In review of literature, general oceanography of Arabian Sea and Bay of Bengal and hydrography of eastern Arabian Sea and western Bay of Bengal. It deals with the distribution patterns of primary production, chlorophyll a, phytoplankton composition and particulate organic carbon in the eastern Arabian Sea and western Bay of Bengal during different seasons. Factors that affect primary productivity are irradiance, temperature, stability of the surface waters, nutrients and zooplankton grazing. The differential biological response of eastern Arabian Sea and western Bay of Bengal to monsoonal regimes. A precise estimation on the primary production of the entire EEZ of India on a regional basis and on a seasonal scale would be the only way to achieve any kind of predictive assessment on the fish stock and their sustainable yield. This study mainly envisages the qualitative and quantitative aspects on the magnitude of phytoplankton standing crop and production of organic carbon and their relationship to environmental characteristics during summer monsoon, Inter monsoon and winter monsoon periods in the east and west coasts of the Indian EEZ.This study revealed that the seasonality exerts a great impact on the biological production in the eastern Arabian Sea and western Bay of Bengal. High biological production may be the reason why most of the fish landings are Concentrated in the west coast of India than east coast. The present data on Phytoplankton production rate and the species composition will provide a meaningful ground for evaluations of exploitable renewable resources of the IndianEEZ

Development of Shape Descriptors Based on Legendre Polynomials and Content-Based Retrieval of Scoliosis Images

The wealth of information available freely on the web and medical image databases poses a major problem for the end users: how to find the information needed? Content –Based Image Retrieval is the obvious solution.A standard called MPEG-7 was evolved to address the interoperability issues of content-based search.The work presented in this thesis mainly concentrates on developing new shape descriptors and a framework for content – based retrieval of scoliosis images.New region-based and contour based shape descriptor is developed based on orthogonal Legendre polymomials.A novel system for indexing and retrieval of digital spine radiographs with scoliosis is presented.

Studies On The Effect Of Laser Radiation And Other Mutagens On Plants

The effect of lasers of three wavelengths in the visible region - 476, 488 and 514 nm on mitotic and meiotic cell divisions, growth, yield and activity of specific enzymes were studied in two taxonomically diverse plant species — A/lium cepa L. and Vicia faba. The effect of laser exposures was compared with the effect of two physical mutagens (Gamma and Ultraviolet radiations) and two chemical mutagens (Ethyl Methane Sulphonate and Hydroxyl amine). The study indicated that lasers could be mutagenic causing aberration in the mitotic and meiotic cell divisions while also producing changes in the growth and yield of the plants. Lasers of higher wavelengths 488 and 514 nm caused aberrations in the early stages of mitotic cell division whereas lasers of lower wavelengths (476 nm) caused more aberrations in the later stages of mitotic cell division. Laser exposure of 488 nm wavelength at power density 400 mW induced higher mitotic and meiotic aberrations and also induced higher pollen sterility than lasers of 476 and 514 nm. The frequency of mitotic aberrations induced by lasers was lesser than that caused by y-irradiation but comparable to that induced by EMS and HA. Lasers cause mutations in higher frequencies than UV. Lasers had a stimulatory effect on growth and yield in both plant species. This stimulatory effect of lasers on germination could not however be correlated to the activity of amylase and protease, the key enzymes in seed gennination. Enzymes such as peroxidase and catalase, involved in scavenging of free oxygen radicals often produced by irradiation, did not show increased activity in laser irradiated samples. Further studies are required for elucidating the exact mechanisms by which lasers cause mutations

A Literature Review on the Benefits of Ethnic and Minority Entrepreneurship

Ethnic/Minority entrepreneur is a business owner who does not represent the majority population of a country. Ethnic/Minority entrepreneurship can bring benefits to individual, society, region, economy, and globe too. So, understanding the importance of ethnic/minority entrepreneurship will really be useful to all the stakeholders. The unique culture and value system of ethnic minorities are the most wanted characteristics for any successful entrepreneur in general. Many industrial nations like U.S, U.K., Germany, etc. have utilised the ethnic minorities to build their economy. The future of the minority businesses looks bright as the world economy is booming, availability of experienced and already successful minority entrepreneurs as role models, and institutional support services. In this paper, literature relating to the benefits of ethnic and minority entrepreneurship is reviewed to understand its magnitude of benefits.

Detection of matchings in a sequence of underwater images through texture analysis

This paper presents an approach to ameliorate the reliability of the correspondence points relating two consecutive images of a sequence. The images are especially difficult to handle, since they have been acquired by a camera looking at the sea floor while carried by an underwater robot. Underwater images are usually difficult to process due to light absorption, changing image radiance and lack of well-defined features. A new approach based on gray-level region matching and selective texture analysis significantly improves the matching reliability

Plan de negocios para la exportación de tarjetas de felicitación e internacionalización de la empresa alternativas expresión

Este trabajo muestra el plan de internacionalización de la empresa Alternativas Expresión, la cual está dedicada a la fabricación y comercialización de productos de expresión social como tarjetas, chocolates y empaques. De esta forma, el estudio se enfocó principalmente en la exportación de Tarjetas de Felicitación, se desarrollaron varias metodologías como el análisis del potencial exportador de la empresa y el estudio de la competencia nacional e internacional. Posteriormente se evaluaron variables macro económicas, comercio internacional y de selección, en once países del norte, centro y sur de América, de los cuales se seleccionaron tres mercados Estados Unidos, Chile y Panamá. Después de la fase de selección de mercados, se elaboró una descripción detallada de cada país seleccionado y se procedió a realizar el trabajo de campo para el país potencial, Estados Unidos. En donde se consideraron diferentes factores como demanda, producto, precio, canales de distribución, competencia, y distribución física internacional. Finalmente, se propuso un plan de mejoramiento para la empresa Alternativas Expresión, con el propósito de minimizar las debilidades que se identificaron en el estudio del potencial exportador y maximizar las oportunidades detalladas en el análisis de la competencia y el trabajo de campo en Estados Unidos.

Actualización perfil de mercado de la Unión Europea

El Proyecto Actualización perfil de mercado unión europea, abarca el estudio de los perfiles de mercado de los 28 países (tras la adición de Croacia el 1 de julio del 2013) miembros de la unión europea (EU), que permita desarrollar una guía de análisis de mercado, por producto y por demanda; ayudando a empresarios colombianos e investigadores de mercados a mejorar la toma de decisiones dentro del intercambio comercial bilateral y crear un valor agregado y mayor desarrollo en la industria. Los empresarios colombianos, independientemente de su tamaño (grandes o PYMES) se pueden valer de la información recogida para evaluar su situación actual y la de los países miembros de la UE y formular estrategias de re-diseño o de incursión de mercado, haciendo posible identificar si sus decisiones son correctas y en caso contrario identificar soluciones y posibles cambios. A raíz del aumento de las relaciones comerciales entre Colombia y los diferentes bloques comerciales, es de vital importancia crear herramientas que aumenten las posibilidades de éxito a los empresarios nacionales y mejoren la relación entre ambas partes; en muchos casos los exportadores colombianos cuentan con bastante información acerca de los procesos necesarios para realizar una exportación e inclusive de cuales países son probablemente los más adecuados para llevar el producto; pero no tienen información detallada, por país, región, producto, y el comportamiento comercial de cada uno; objetivo final es eliminar esa incertidumbre y generar un mayor número de procesos de exportación exitosos y sostenibles.

Aprendre colcant. Projecte de mobilitat de l’IES Felanitx

Este proyecto fue galardonado con el 1er Premio de Innovación Educativa 2009.

Good-visibility computation using graphics hardware

Aquesta tesi tracta del disseny, implementació i discussió d'algoritmes per resoldre problemes de visibilitat i bona-visibilitat utilitzant el hardware gràfic de l'ordinador. Concretament, s'obté una discretització dels mapes de multi-visibilitat i bona-visibilitat a partir d'un conjunt d'objectes de visió i un conjunt d'obstacles. Aquests algoritmes són útils tant per fer càlculs en dues dimensions com en tres dimensions. Fins i tot ens permeten calcular-los sobre terrenys.

Les sHsps en surera: Estudis de funcionalitat

Aquesta tesi es centra en la caracterització funcional d'una proteïna de xoc de calor de baix pes molecular (Small Heat Shock Protein - sHSP) de classe I de surera pel que fa a la seva capacitat per protegir les cèl·lules de l'estrès i per estabilitzar les membranes. Les sHsps són proteïnes que s'expressen en condicions d'estrès cel·lular. Encara que certs aspectes funcionals de les sHsps són ben coneguts, el nostre treball aporta informacions noves sobre el paper de les diferents regions de la proteïna, especialment de la regió N-terminal. L'objectiu concret d'aquest treball és determinar la funció termoprotectora de QsHsp17.4-CI, una sHsp de classe I oobtinguda a partir de les cèl·lules de fel·lema d'alzina surera, en un model bacterià i analitzar la importància de les diferents regions de la proteïna en aquesta funció. Amb aquesta finalitat s'han dissenyat dues proteïnes parcials derivades de QsHsp17.4-CI: una a la que li falta la regió N-terminal (C105) i una altra amb pràcticament tot el domini -cristal·lí deleccionat (N61), i una tercera, derivada de QsHs10-CI, a la que li falta la meitat del domini -cristal·lí (Hsp10). També s'estudia la possible capacitat estabilitzadora de membranes i la capacitat de modificar l'expressió d'altres Hsps quan s'expressa de forma heteròloga. Els nostres resultats demostren que l'expressió de QsHsp17.4-CI protegeix a les cèl·lules d'E.coli de l'estrès tèrmic alhora que la regió N-terminal i la regió consens II del domini -cristal·lí són imprescindibles per aquesta funció de protecció. En relació a un possible paper en les membranes, els estudis de localització subcel·lular mostren que QsHsp17.4-CI colocalitza amb la fracció membranes i que la regió N-terminal de la proteïna és responsable d'aquesta colocalització. No s'ha pogut demostrar, però, que la localització amb la membrana estigui associada a un efecte protector d'aquesta: en cap cas la sobrexpressió de les proteïnes modifica la composició d'àcids grassos i només N61, que no té acció termoprotectora, altera l'estat fisico-químic de la membrana. En estudis d'expressió de novo en E.coli s'ha observat que, a diferència de les altres proteïnes heteròlogues, N61 activa l'expressió de la majoria de Hsps d'E.coli fent pensar en una possible relació entre l'estat físic de la membrana i l'activació de la resposta a l'estrès. En resum, en aquest treball hem provat la capacitat protectora de QsHsp17.4 i aportem noves dades sobre la importància de la regió N-terminal i la regió consens II del domini -cristal·lí en aquesta funció. Per altra banda, es suggereix que QsHsp17.4 podria interaccionar amb la membrana d'E.coli i que la regió N-terminal seria imprescindible per aquesta interacció. Finalment hem determinat que les proteïnes que provoquen variacions en l'estat de fluïdesa de la membrana poden activar la resposta al xoc de calor per part de la cèl·lula bacteriana.

Incorporating Legacy Soil pH Databases into Digital Soil Maps

Soil data and reliable soil maps are imperative for environmental management. conservation and policy. Data from historical point surveys, e.g. experiment site data and farmers fields can serve this purpose. However, legacy soil information is not necessarily collected for spatial analysis and mapping such that the data may not have immediately useful geo-references. Methods are required to utilise these historical soil databases so that we can produce quantitative maps of soil propel-ties to assess spatial and temporal trends but also to assess where future sampling is required. This paper discusses two such databases: the Representative Soil Sampling Scheme which has monitored the agricultural soil in England and Wales from 1969 to 2003 (between 400 and 900 bulked soil samples were taken annually from different agricultural fields); and the former State Chemistry Laboratory, Victoria, Australia where between 1973 and 1994 approximately 80,000 soil samples were submitted for analysis by farmers. Previous statistical analyses have been performed using administrative regions (with sharp boundaries) for both databases, which are largely unrelated to natural features. For a more detailed spatial analysis that call be linked to climate and terrain attributes, gradual variation of these soil properties should be described. Geostatistical techniques such as ordinary kriging are suited to this. This paper describes the format of the databases and initial approaches as to how they can be used for digital soil mapping. For this paper we have selected soil pH to illustrate the analyses for both databases.

Uses of airborne scanning laser altimetry for river flood modelling and coastal geomorphology

Two ongoing projects at ESSC that involve the development of new techniques for extracting information from airborne LiDAR data and combining this information with environmental models will be discussed. The first project in conjunction with Bristol University is aiming to improve 2-D river flood flow models by using remote sensing to provide distributed data for model calibration and validation. Airborne LiDAR can provide such models with a dense and accurate floodplain topography together with vegetation heights for parameterisation of model friction. The vegetation height data can be used to specify a friction factor at each node of a model’s finite element mesh. A LiDAR range image segmenter has been developed which converts a LiDAR image into separate raster maps of surface topography and vegetation height for use in the model. Satellite and airborne SAR data have been used to measure flood extent remotely in order to validate the modelled flood extent. Methods have also been developed for improving the models by decomposing the model’s finite element mesh to reflect floodplain features such as hedges and trees having different frictional properties to their surroundings. Originally developed for rural floodplains, the segmenter is currently being extended to provide DEMs and friction parameter maps for urban floods, by fusing the LiDAR data with digital map data. The second project is concerned with the extraction of tidal channel networks from LiDAR. These networks are important features of the inter-tidal zone, and play a key role in tidal propagation and in the evolution of salt-marshes and tidal flats. The study of their morphology is currently an active area of research, and a number of theories related to networks have been developed which require validation using dense and extensive observations of network forms and cross-sections. The conventional method of measuring networks is cumbersome and subjective, involving manual digitisation of aerial photographs in conjunction with field measurement of channel depths and widths for selected parts of the network. A semi-automatic technique has been developed to extract networks from LiDAR data of the inter-tidal zone. A multi-level knowledge-based approach has been implemented, whereby low level algorithms first extract channel fragments based mainly on image properties then a high level processing stage improves the network using domain knowledge. The approach adopted at low level uses multi-scale edge detection to detect channel edges, then associates adjacent anti-parallel edges together to form channels. The higher level processing includes a channel repair mechanism.

Introduction: perspectives for economically important wild species and neglected crops in the Mediterranean

The Mediterranean region is one of the major centres of origin and diversification of cultivated plants and many crop wild relatives are found there. In addition, many native species are still widely harvested from the wild for food, medicine and other uses and some of these have potential for development as alternative crop especially in marginal zones. While there have been several recent initiatives that address the cataloguing and conservation of these species, such as the Network on Identification, Conservation and Use of Wild Plants in the Mediterranean Region (MEDUSA and the Bioversity International (IPGRI) studies on Underutilized Mediterranean Species (VMS), no comprehensive assessment has yet been made and little work undertaken on their agricultural potential. It has been confidently predicted that consequences of global change in the Mediterranean region - population movements and migrations, changes in disturbance regimes, and climate change - will be serious. One the one hand, this will affect the survival prospects of many of these underutilized species and on the other hand it will enhance their importance as the source of potential new crop germplasm. The conservation and availability of genetic diversity of both crops and underutilized species is essential if we are to be able to meet the increasing demand for food and other crops that will be adapted to the new ecoclimatic envelopes that will develop in the region as a consequence of global change. The rapid rate of climatic and other change that is expected adds urgency to the task of assessing, conserving and sustainably using this rich diversity of wild species of economic value in the region but new strategies will be need to be developed to achieve this. The Mediterranean region has the potential of becoming a major source of new crop development in the coming decades.