Deep learning of representations and its application to computer vision

L’objectif de cette thèse par articles est de présenter modestement quelques étapes du parcours qui mènera (on espère) à une solution générale du problème de l’intelligence artificielle. Cette thèse contient quatre articles qui présentent chacun une différente nouvelle méthode d’inférence perceptive en utilisant l’apprentissage machine et, plus particulièrement, les réseaux neuronaux profonds. Chacun de ces documents met en évidence l’utilité de sa méthode proposée dans le cadre d’une tâche de vision par ordinateur. Ces méthodes sont applicables dans un contexte plus général, et dans certains cas elles on tété appliquées ailleurs, mais ceci ne sera pas abordé dans le contexte de cette de thèse. Dans le premier article, nous présentons deux nouveaux algorithmes d’inférence variationelle pour le modèle génératif d’images appelé codage parcimonieux “spike- and-slab” (CPSS). Ces méthodes d’inférence plus rapides nous permettent d’utiliser des modèles CPSS de tailles beaucoup plus grandes qu’auparavant. Nous démontrons qu’elles sont meilleures pour extraire des détecteur de caractéristiques quand très peu d’exemples étiquetés sont disponibles pour l’entraînement. Partant d’un modèle CPSS, nous construisons ensuite une architecture profonde, la machine de Boltzmann profonde partiellement dirigée (MBP-PD). Ce modèle a été conçu de manière à simplifier d’entraînement des machines de Boltzmann profondes qui nécessitent normalement une phase de pré-entraînement glouton pour chaque couche. Ce problème est réglé dans une certaine mesure, mais le coût d’inférence dans le nouveau modèle est relativement trop élevé pour permettre de l’utiliser de manière pratique. Dans le deuxième article, nous revenons au problème d’entraînement joint de machines de Boltzmann profondes. Cette fois, au lieu de changer de famille de modèles, nous introduisons un nouveau critère d’entraînement qui donne naissance aux machines de Boltzmann profondes à multiples prédictions (MBP-MP). Les MBP-MP sont entraînables en une seule étape et ont un meilleur taux de succès en classification que les MBP classiques. Elles s’entraînent aussi avec des méthodes variationelles standard au lieu de nécessiter un classificateur discriminant pour obtenir un bon taux de succès en classification. Par contre, un des inconvénients de tels modèles est leur incapacité de générer deséchantillons, mais ceci n’est pas trop grave puisque la performance de classification des machines de Boltzmann profondes n’est plus une priorité étant donné les dernières avancées en apprentissage supervisé. Malgré cela, les MBP-MP demeurent intéressantes parce qu’elles sont capable d’accomplir certaines tâches que des modèles purement supervisés ne peuvent pas faire, telles que celle de classifier des données incomplètes ou encore celle de combler intelligemment l’information manquante dans ces données incomplètes. Le travail présenté dans cette thèse s’est déroulé au milieu d’une période de transformations importantes du domaine de l’apprentissage à réseaux neuronaux profonds qui a été déclenchée par la découverte de l’algorithme de “dropout” par Geoffrey Hinton. Dropout rend possible un entraînement purement supervisé d’architectures de propagation unidirectionnel sans être exposé au danger de sur- entraînement. Le troisième article présenté dans cette thèse introduit une nouvelle fonction d’activation spécialement con ̧cue pour aller avec l’algorithme de Dropout. Cette fonction d’activation, appelée maxout, permet l’utilisation de aggrégation multi-canal dans un contexte d’apprentissage purement supervisé. Nous démontrons comment plusieurs tâches de reconnaissance d’objets sont mieux accomplies par l’utilisation de maxout. Pour terminer, sont présentons un vrai cas d’utilisation dans l’industrie pour la transcription d’adresses de maisons à plusieurs chiffres. En combinant maxout avec une nouvelle sorte de couche de sortie pour des réseaux neuronaux de convolution, nous démontrons qu’il est possible d’atteindre un taux de succès comparable à celui des humains sur un ensemble de données coriace constitué de photos prises par les voitures de Google. Ce système a été déployé avec succès chez Google pour lire environ cent million d’adresses de maisons.

Changing Role of Academic Libraries in the E-learning Environment: Issues and Challenges

The development of computer and network technology is changing the education scenario and transforming the teaching and learning process from the traditional physical environment to the digital environment. It is now possible to access vast amount of information online and enable one to one communication without the confines of place or time. While E-learning and teaching is unlikely to replace face-to-face training and education it is becoming an additional delivery method, providing new learning opportunities to many users. It is also causing an impact on library services as the increased use of ICT and web based learning technologies have paved the way for providing new ICT based services and resources to the users. Online learning has a crucial role in user education, information literacy programmes and in training the library professionals. It can help students become active learners, and libraries will have to play a greater role in this process of transformation. The significance of libraries within an institution has improved due to the fact that academic libraries and information services are now responsible for e-learning within their organization.

A User-Centric Perspective on Parallel Programming with Focus on OpenMP

The process of developing software that takes advantage of multiple processors is commonly referred to as parallel programming. For various reasons, this process is much harder than the sequential case. For decades, parallel programming has been a problem for a small niche only: engineers working on parallelizing mostly numerical applications in High Performance Computing. This has changed with the advent of multi-core processors in mainstream computer architectures. Parallel programming in our days becomes a problem for a much larger group of developers. The main objective of this thesis was to find ways to make parallel programming easier for them. Different aims were identified in order to reach the objective: research the state of the art of parallel programming today, improve the education of software developers about the topic, and provide programmers with powerful abstractions to make their work easier. To reach these aims, several key steps were taken. To start with, a survey was conducted among parallel programmers to find out about the state of the art. More than 250 people participated, yielding results about the parallel programming systems and languages in use, as well as about common problems with these systems. Furthermore, a study was conducted in university classes on parallel programming. It resulted in a list of frequently made mistakes that were analyzed and used to create a programmers' checklist to avoid them in the future. For programmers' education, an online resource was setup to collect experiences and knowledge in the field of parallel programming - called the Parawiki. Another key step in this direction was the creation of the Thinking Parallel weblog, where more than 50.000 readers to date have read essays on the topic. For the third aim (powerful abstractions), it was decided to concentrate on one parallel programming system: OpenMP. Its ease of use and high level of abstraction were the most important reasons for this decision. Two different research directions were pursued. The first one resulted in a parallel library called AthenaMP. It contains so-called generic components, derived from design patterns for parallel programming. These include functionality to enhance the locks provided by OpenMP, to perform operations on large amounts of data (data-parallel programming), and to enable the implementation of irregular algorithms using task pools. AthenaMP itself serves a triple role: the components are well-documented and can be used directly in programs, it enables developers to study the source code and learn from it, and it is possible for compiler writers to use it as a testing ground for their OpenMP compilers. The second research direction was targeted at changing the OpenMP specification to make the system more powerful. The main contributions here were a proposal to enable thread-cancellation and a proposal to avoid busy waiting. Both were implemented in a research compiler, shown to be useful in example applications, and proposed to the OpenMP Language Committee.

User Profile Management and Selection in Context-Aware Service Platforms for Networks Beyond 3G

In recent years, progress in the area of mobile telecommunications has changed our way of life, in the private as well as the business domain. Mobile and wireless networks have ever increasing bit rates, mobile network operators provide more and more services, and at the same time costs for the usage of mobile services and bit rates are decreasing. However, mobile services today still lack functions that seamlessly integrate into users’ everyday life. That is, service attributes such as context-awareness and personalisation are often either proprietary, limited or not available at all. In order to overcome this deficiency, telecommunications companies are heavily engaged in the research and development of service platforms for networks beyond 3G for the provisioning of innovative mobile services. These service platforms are to support such service attributes. Service platforms are to provide basic service-independent functions such as billing, identity management, context management, user profile management, etc. Instead of developing own solutions, developers of end-user services such as innovative messaging services or location-based services can utilise the platform-side functions for their own purposes. In doing so, the platform-side support for such functions takes away complexity, development time and development costs from service developers. Context-awareness and personalisation are two of the most important aspects of service platforms in telecommunications environments. The combination of context-awareness and personalisation features can also be described as situation-dependent personalisation of services. The support for this feature requires several processing steps. The focus of this doctoral thesis is on the processing step, in which the user’s current context is matched against situation-dependent user preferences to find the matching user preferences for the current user’s situation. However, to achieve this, a user profile management system and corresponding functionality is required. These parts are also covered by this thesis. Altogether, this thesis provides the following contributions: The first part of the contribution is mainly architecture-oriented. First and foremost, we provide a user profile management system that addresses the specific requirements of service platforms in telecommunications environments. In particular, the user profile management system has to deal with situation-specific user preferences and with user information for various services. In order to structure the user information, we also propose a user profile structure and the corresponding user profile ontology as part of an ontology infrastructure in a service platform. The second part of the contribution is the selection mechanism for finding matching situation-dependent user preferences for the personalisation of services. This functionality is provided as a sub-module of the user profile management system. Contrary to existing solutions, our selection mechanism is based on ontology reasoning. This mechanism is evaluated in terms of runtime performance and in terms of supported functionality compared to other approaches. The results of the evaluation show the benefits and the drawbacks of ontology modelling and ontology reasoning in practical applications.

An algorithmic approach for collaborative-based prediction of user contexts in ubiquitous environments under consideration of legal implications

Mit Hilfe der Vorhersage von Kontexten können z. B. Dienste innerhalb einer ubiquitären Umgebung proaktiv an die Bedürfnisse der Nutzer angepasst werden. Aus diesem Grund hat die Kontextvorhersage einen signifikanten Stellenwert innerhalb des ’ubiquitous computing’. Nach unserem besten Wissen, verwenden gängige Ansätze in der Kontextvorhersage ausschließlich die Kontexthistorie des Nutzers als Datenbasis, dessen Kontexte vorhersagt werden sollen. Im Falle, dass ein Nutzer unerwartet seine gewohnte Verhaltensweise ändert, enthält die Kontexthistorie des Nutzers keine geeigneten Informationen, um eine zuverlässige Kontextvorhersage zu gewährleisten. Daraus folgt, dass Vorhersageansätze, die ausschließlich die Kontexthistorie des Nutzers verwenden, dessen Kontexte vorhergesagt werden sollen, fehlschlagen könnten. Um die Lücke der fehlenden Kontextinformationen in der Kontexthistorie des Nutzers zu schließen, führen wir den Ansatz zur kollaborativen Kontextvorhersage (CCP) ein. Dabei nutzt CCP bestehende direkte und indirekte Relationen, die zwischen den Kontexthistorien der verschiedenen Nutzer existieren können, aus. CCP basiert auf der Singulärwertzerlegung höherer Ordnung, die bereits erfolgreich in bestehenden Empfehlungssystemen eingesetzt wurde. Um Aussagen über die Vorhersagegenauigkeit des CCP Ansatzes treffen zu können, wird dieser in drei verschiedenen Experimenten evaluiert. Die erzielten Vorhersagegenauigkeiten werden mit denen von drei bekannten Kontextvorhersageansätzen, dem ’Alignment’ Ansatz, dem ’StatePredictor’ und dem ’ActiveLeZi’ Vorhersageansatz, verglichen. In allen drei Experimenten werden als Evaluationsbasis kollaborative Datensätze verwendet. Anschließend wird der CCP Ansatz auf einen realen kollaborativen Anwendungsfall, den proaktiven Schutz von Fußgängern, angewendet. Dabei werden durch die Verwendung der kollaborativen Kontextvorhersage Fußgänger frühzeitig erkannt, die potentiell Gefahr laufen, mit einem sich nähernden Auto zu kollidieren. Als kollaborative Datenbasis werden reale Bewegungskontexte der Fußgänger verwendet. Die Bewegungskontexte werden mittels Smartphones, welche die Fußgänger in ihrer Hosentasche tragen, gesammelt. Aus dem Grund, dass Kontextvorhersageansätze in erster Linie personenbezogene Kontexte wie z.B. Standortdaten oder Verhaltensmuster der Nutzer als Datenbasis zur Vorhersage verwenden, werden rechtliche Evaluationskriterien aus dem Recht des Nutzers auf informationelle Selbstbestimmung abgeleitet. Basierend auf den abgeleiteten Evaluationskriterien, werden der CCP Ansatz und weitere bekannte kontextvorhersagende Ansätze bezüglich ihrer Rechtsverträglichkeit untersucht. Die Evaluationsergebnisse zeigen die rechtliche Kompatibilität der untersuchten Vorhersageansätze bezüglich des Rechtes des Nutzers auf informationelle Selbstbestimmung auf. Zum Schluss wird in der Dissertation ein Ansatz für die verteilte und kollaborative Vorhersage von Kontexten vorgestellt. Mit Hilfe des Ansatzes wird eine Möglichkeit aufgezeigt, um den identifizierten rechtlichen Probleme, die bei der Vorhersage von Kontexten und besonders bei der kollaborativen Vorhersage von Kontexten, entgegenzuwirken.

The Human in the Loop: User Participation in Self-Adaptive Software

Self-adaptive software provides a profound solution for adapting applications to changing contexts in dynamic and heterogeneous environments. Having emerged from Autonomic Computing, it incorporates fully autonomous decision making based on predefined structural and behavioural models. The most common approach for architectural runtime adaptation is the MAPE-K adaptation loop implementing an external adaptation manager without manual user control. However, it has turned out that adaptation behaviour lacks acceptance if it does not correspond to a user’s expectations – particularly for Ubiquitous Computing scenarios with user interaction. Adaptations can be irritating and distracting if they are not appropriate for a certain situation. In general, uncertainty during development and at run-time causes problems with users being outside the adaptation loop. In a literature study, we analyse publications about self-adaptive software research. The results show a discrepancy between the motivated application domains, the maturity of examples, and the quality of evaluations on the one hand and the provided solutions on the other hand. Only few publications analysed the impact of their work on the user, but many employ user-oriented examples for motivation and demonstration. To incorporate the user within the adaptation loop and to deal with uncertainty, our proposed solutions enable user participation for interactive selfadaptive software while at the same time maintaining the benefits of intelligent autonomous behaviour. We define three dimensions of user participation, namely temporal, behavioural, and structural user participation. This dissertation contributes solutions for user participation in the temporal and behavioural dimension. The temporal dimension addresses the moment of adaptation which is classically determined by the self-adaptive system. We provide mechanisms allowing users to influence or to define the moment of adaptation. With our solution, users can have full control over the moment of adaptation or the self-adaptive software considers the user’s situation more appropriately. The behavioural dimension addresses the actual adaptation logic and the resulting run-time behaviour. Application behaviour is established during development and does not necessarily match the run-time expectations. Our contributions are three distinct solutions which allow users to make changes to the application’s runtime behaviour: dynamic utility functions, fuzzy-based reasoning, and learning-based reasoning. The foundation of our work is a notification and feedback solution that improves intelligibility and controllability of self-adaptive applications by implementing a bi-directional communication between self-adaptive software and the user. The different mechanisms from the temporal and behavioural participation dimension require the notification and feedback solution to inform users on adaptation actions and to provide a mechanism to influence adaptations. Case studies show the feasibility of the developed solutions. Moreover, an extensive user study with 62 participants was conducted to evaluate the impact of notifications before and after adaptations. Although the study revealed that there is no preference for a particular notification design, participants clearly appreciated intelligibility and controllability over autonomous adaptations.

Validating Co-Training Models for Web Image Classification

Co-training is a semi-supervised learning method that is designed to take advantage of the redundancy that is present when the object to be identified has multiple descriptions. Co-training is known to work well when the multiple descriptions are conditional independent given the class of the object. The presence of multiple descriptions of objects in the form of text, images, audio and video in multimedia applications appears to provide redundancy in the form that may be suitable for co-training. In this paper, we investigate the suitability of utilizing text and image data from the Web for co-training. We perform measurements to find indications of conditional independence in the texts and images obtained from the Web. Our measurements suggest that conditional independence is likely to be present in the data. Our experiments, within a relevance feedback framework to test whether a method that exploits the conditional independence outperforms methods that do not, also indicate that better performance can indeed be obtained by designing algorithms that exploit this form of the redundancy when it is present.

Adaption of Courses and Learning Environment to the User Context in dotLRN

The explosive growth of Internet during the last years has been reflected in the ever-increasing amount of the diversity and heterogeneity of user preferences, types and features of devices and access networks. Usually the heterogeneity in the context of the users which request Web contents is not taken into account by the servers that deliver them implying that these contents will not always suit their needs. In the particular case of e-learning platforms this issue is especially critical due to the fact that it puts at stake the knowledge acquired by their users. In the following paper we present a system that aims to provide the dotLRN e-learning platform with the capability to adapt to its users context. By integrating dotLRN with a multi-agent hypermedia system, online courses being undertaken by students as well as their learning environment are adapted in real time

Curriculum sequencing for an e-learning system based on learning styles

This work shows the use of adaptation techniques involved in an e-learning system that considers students' learning styles and students' knowledge states. The mentioned e-learning system is built on a multiagent framework designed to examine opportunities to improve the teaching and to motivate the students to learn what they want in a user-friendly and assisted environment

Desarrollo de una aplicación de computador para analizar parámetros estadísticos genético-poblacionales de microsatélites en poblaciones humanas

La información y los datos genéticos que emanan hoy de las investigaciones del genoma humano demandan el desarrollo de herramientas informáticas capaces de procesar la gran cantidad de información disponible. La mayor cantidad de datos genéticos es el resultado de equipos que realizan el análisis simultáneo de cientos o miles de polimorfismos o variaciones genéticas, de nuevas técnicas de laboratorio de mayor rendimiento que, en conjunto, ofrecen una mayor disponibilidad de información en un corto espacio de tiempo. Esta problemática conduce a la necesidad de desarrollar nuevas herramientas informáticas capaces de lidiar con este mayor volumen de datos genéticos. En el caso de la genética de poblaciones, a pesar de que existen herramientas informáticas que permiten procesar y facilitar el análisis de los datos, estas tienen limitaciones como la falta de conocimiento de los usuarios de algunos lenguajes de programación para alimentar la información y otras herramientas informáticas no realizan todas las estimaciones que se requieren y otros presentan limitaciones en cuanto al número de datos que pueden incorporar o manejar. En algunos casos hay redundancia al tener que usarse dos o más herramientas para poder procesar un conjunto de datos de información genética. El presente trabajo tiene por objetivo el desarrollo de una herramienta informática basada en aplicaciones de computador comunes, en este caso Microsoft Excel® y que resuelva todos los problemas y las limitaciones descritas antes. El desarrollo del conjunto de subprogramas que constituyen a Lustro; permiten superar lo anterior, presentar los resultados en un ambiente sencillo, conocido y fácil de operar, simplificando de esta forma el proceso de adaptación del usuario del programa, sin entrenamiento previo, obteniéndose en corto tiempo el procesamiento de la información genética de interés.

IT and other training in the always connected era

Abstract Google and YouTube are quickly becoming the training resource of choice for the IT literate, especially in relation to computer based applications. Many businesses are addressing this training issue in a number of ways, some more successful than others. Find out what the IT services at the university are doing to adapt to this change and contribute to the discussion on how the approach could be improved. Before the talk you could have a look at the following; * One service that has been licenced is Lynda http://go.soton.ac.uk/lynda or lynda.com (note you have to enter www.southampton.ac.uk as the organisation if you don’t log in through the go.soton link) * The IT training team publish a portfolio of systems and courses at http://www.southampton.ac.uk/isolutions/computing/training/portfolio/index.php. * More and more internal systems are being supported through online guides such as http://go.soton.ac.uk/bgsg

User-Centred Methods for Measuring the Value of Open Data

A project to identify metrics for assessing the quality of open data based on the needs of small voluntary sector organisations in the UK and India. For this project we assumed the purpose of open data metrics is to determine the value of a group of open datasets to a defined community of users. We adopted a much more user-centred approach than most open data research using small structured workshops to identify users’ key problems and then working from those problems to understand how open data can help address them and the key attributes of the data if it is to be successful. We then piloted different metrics that might be used to measure the presence of those attributes. The result was six metrics that we assessed for validity, reliability, discrimination, transferability and comparability. This user-centred approach to open data research highlighted some fundamental issues with expanding the use of open data from its enthusiast base.

An automatic correction tool for relational database schemas

A Web-based tool developed to automatically correct relational database schemas is presented. This tool has been integrated into a more general e-learning platform and is used to reinforce teaching and learning on database courses. This platform assigns to each student a set of database problems selected from a common repository. The student has to design a relational database schema and enter it into the system through a user friendly interface specifically designed for it. The correction tool corrects the design and shows detected errors. The student has the chance to correct them and send a new solution. These steps can be repeated as many times as required until a correct solution is obtained. Currently, this system is being used in different introductory database courses at the University of Girona with very promising results

Dental training system using haptic technology

In the U.K., dental students require to perform training and practice on real human tissues at the very early stage of their courses. Currently, the human tissues, such as decayed teeth, are mounted in a human head like physical model. The problems with these models in teaching are; (1) every student operates on tooth, which are always unique; (2) the process cannot be recorded for examination purposes and (3) same training are not repeatable. The aim of the PHATOM Project is to develop a dental training system using Haptic technology. This paper documents the project background, specification, research and development of the first prototype system. It also discusses the research in the visual display, haptic devices and haptic rendering. This includes stereo vision, motion parallax, volumetric modelling, surface remapping algorithms as well as analysis design of the system. A new volumetric to surface model transformation algorithm is also introduced. This paper includes the future work on the system development and research.

Inferring population history with DIY ABC: a user-friendly approach to approximate Bayesian computation

Genetic data obtained on population samples convey information about their evolutionary history. Inference methods can extract part of this information but they require sophisticated statistical techniques that have been made available to the biologist community (through computer programs) only for simple and standard situations typically involving a small number of samples. We propose here a computer program (DIY ABC) for inference based on approximate Bayesian computation (ABC), in which scenarios can be customized by the user to fit many complex situations involving any number of populations and samples. Such scenarios involve any combination of population divergences, admixtures and population size changes. DIY ABC can be used to compare competing scenarios, estimate parameters for one or more scenarios and compute bias and precision measures for a given scenario and known values of parameters (the current version applies to unlinked microsatellite data). This article describes key methods used in the program and provides its main features. The analysis of one simulated and one real dataset, both with complex evolutionary scenarios, illustrates the main possibilities of DIY ABC.