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.

Reinforcement Learning Approaches to Power System Scheduling

One major component of power system operation is generation scheduling. The objective of the work is to develop efficient control strategies to the power scheduling problems through Reinforcement Learning approaches. The three important active power scheduling problems are Unit Commitment, Economic Dispatch and Automatic Generation Control. Numerical solution methods proposed for solution of power scheduling are insufficient in handling large and complex systems. Soft Computing methods like Simulated Annealing, Evolutionary Programming etc., are efficient in handling complex cost functions, but find limitation in handling stochastic data existing in a practical system. Also the learning steps are to be repeated for each load demand which increases the computation time.Reinforcement Learning (RL) is a method of learning through interactions with environment. The main advantage of this approach is it does not require a precise mathematical formulation. It can learn either by interacting with the environment or interacting with a simulation model. Several optimization and control problems have been solved through Reinforcement Learning approach. The application of Reinforcement Learning in the field of Power system has been a few. The objective is to introduce and extend Reinforcement Learning approaches for the active power scheduling problems in an implementable manner. The main objectives can be enumerated as:(i) Evolve Reinforcement Learning based solutions to the Unit Commitment Problem.(ii) Find suitable solution strategies through Reinforcement Learning approach for Economic Dispatch. (iii) Extend the Reinforcement Learning solution to Automatic Generation Control with a different perspective. (iv) Check the suitability of the scheduling solutions to one of the existing power systems.First part of the thesis is concerned with the Reinforcement Learning approach to Unit Commitment problem. Unit Commitment Problem is formulated as a multi stage decision process. Q learning solution is developed to obtain the optimwn commitment schedule. Method of state aggregation is used to formulate an efficient solution considering the minimwn up time I down time constraints. The performance of the algorithms are evaluated for different systems and compared with other stochastic methods like Genetic Algorithm.Second stage of the work is concerned with solving Economic Dispatch problem. A simple and straight forward decision making strategy is first proposed in the Learning Automata algorithm. Then to solve the scheduling task of systems with large number of generating units, the problem is formulated as a multi stage decision making task. The solution obtained is extended in order to incorporate the transmission losses in the system. To make the Reinforcement Learning solution more efficient and to handle continuous state space, a fimction approximation strategy is proposed. The performance of the developed algorithms are tested for several standard test cases. Proposed method is compared with other recent methods like Partition Approach Algorithm, Simulated Annealing etc.As the final step of implementing the active power control loops in power system, Automatic Generation Control is also taken into consideration.Reinforcement Learning has already been applied to solve Automatic Generation Control loop. The RL solution is extended to take up the approach of common frequency for all the interconnected areas, more similar to practical systems. Performance of the RL controller is also compared with that of the conventional integral controller.In order to prove the suitability of the proposed methods to practical systems, second plant ofNeyveli Thennal Power Station (NTPS IT) is taken for case study. The perfonnance of the Reinforcement Learning solution is found to be better than the other existing methods, which provide the promising step towards RL based control schemes for practical power industry.Reinforcement Learning is applied to solve the scheduling problems in the power industry and found to give satisfactory perfonnance. Proposed solution provides a scope for getting more profit as the economic schedule is obtained instantaneously. Since Reinforcement Learning method can take the stochastic cost data obtained time to time from a plant, it gives an implementable method. As a further step, with suitable methods to interface with on line data, economic scheduling can be achieved instantaneously in a generation control center. Also power scheduling of systems with different sources such as hydro, thermal etc. can be looked into and Reinforcement Learning solutions can be achieved.

A Reinforcement Learning Algorithm to Economic Dispatch Considering Transmission Losses

Reinforcement Learning (RL) refers to a class of learning algorithms in which learning system learns which action to take in different situations by using a scalar evaluation received from the environment on performing an action. RL has been successfully applied to many multi stage decision making problem (MDP) where in each stage the learning systems decides which action has to be taken. Economic Dispatch (ED) problem is an important scheduling problem in power systems, which decides the amount of generation to be allocated to each generating unit so that the total cost of generation is minimized without violating system constraints. In this paper we formulate economic dispatch problem as a multi stage decision making problem. In this paper, we also develop RL based algorithm to solve the ED problem. The performance of our algorithm is compared with other recent methods. The main advantage of our method is it can learn the schedule for all possible demands simultaneously.

A Reinforcement Learning Approach to Economic Dispatch using Neural Networks

This paper presents a Reinforcement Learning (RL) approach to economic dispatch (ED) using Radial Basis Function neural network. We formulate the ED as an N stage decision making problem. We propose a novel architecture to store Qvalues and present a learning algorithm to learn the weights of the neural network. Even though many stochastic search techniques like simulated annealing, genetic algorithm and evolutionary programming have been applied to ED, they require searching for the optimal solution for each load demand. Also they find limitation in handling stochastic cost functions. In our approach once we learn the Q-values, we can find the dispatch for any load demand. We have recently proposed a RL approach to ED. In that approach, we could find only the optimum dispatch for a set of specified discrete values of power demand. The performance of the proposed algorithm is validated by taking IEEE 6 bus system, considering transmission losses

An introduction to ordinary differential equations by computer algebra-systems

We report on an elementary course in ordinary differential equations (odes) for students in engineering sciences. The course is also intended to become a self-study package for odes and is is based on several interactive computer lessons using REDUCE and MATHEMATICA . The aim of the course is not to do Computer Algebra (CA) by example or to use it for doing classroom examples. The aim ist to teach and to learn mathematics by using CA-systems.

Local food security and principle of organic farming (from farm to fork) in context of food culture in Indonesia: Minangkabau's case study

The basic idea behind improving local food security consists of two paths; first, accessibility (price, stock) and second, availability (quantity and biodiversity); both are perquisites to the provision of nutrients and a continuous food supply with locally available resources. The objectives of this thesis are to investigate if indigenous knowledge still plays an important role in traditional farming in the Minangkabau`s culture, thus supporting local food security. If the indigenous knowledge still plays a role in food culture in the Minangkabau`s culture which is linked to the matrilineal role and leads to a sound nutrition. Further, it should be tested if marantau influences traditional farming and food culture in Minangkabau`s, and if the local government plays a role in changing of traditional farming systems and food culture. Furthermore this thesis wants to prove if education and gender are playing a role in changing traditional farming system and food culture, and if the mass media affects traditional farming systems and food culture for the Minangkabau. The study was completed at four locations in West Sumatera; Nagari Ulakan (NU) (coastal area), Nagari Aia Batumbuak (NAB) (hilly area), Nagari Padang Laweh Malalo (NPLM) (lake area), Nagari Pandai Sikek (NPS) (hilly area). The rainfall ranged from 1400- 4800 mm annually with fertile soils. Data was collected by using PRA (Participatory Rural Appraisal) to investigate indigenous knowledge (IK) and its interactions, which is also combining with in depth-interview, life history, a survey using semi-structured-questionnaire, pictures, mapping, and expert interview. The data was collected from June - September 2009 and June 2010. The materials are; map of area, list of names, questionnaires, voices recorder, note book, and digital camera. The sampling method was snowball sampling which resulted in the qualitative and quantitative data taken. For qualitative data, ethnography and life history was used. For quantitative, a statistical survey with a semi-structured questionnaire was used. 50 respondents per each site participated voluntarily. Data was analyzed by performing MAXQDA 10, and F4 audio analysis software (created and developed by Philip-University Marburg). The data is clustered based on causality. The results show that; the role of IK on TFS (traditional farming system) shown on NPLM which has higher food crop biodiversity in comparison to the other three places even though it has relatively similar temperature and rainfall. This high food crop biodiversity is due to the awareness of local people who realized that they lived in unfavourable climate and topography; therefore they are more prepared for any changes that may occur. Carbohydrate intake is 100 % through rice even though they are growing different staple crops. Whereas most of the people said in the interviews that not eating rice is like not really eating for them. In addition to that, mothers still play an important role in kitchen activities. But when the agriculture income is low, mothers have to decide whether to change the meals or to feel insecure about their food supply. Marantau yields positive impact through the remittances it provides to invest on the farm. On the other hand, it results in fewer workers for agriculture, and therefore a negative impact on the transfer of IK. The investigation showed that the local government has a PTS (Padi Tanam Sabatang) programme which still does not guarantee that the farmers are getting sufficient revenue from their land. The low agricultural income leads to situation of potential food insecurity. It is evident that education is equal among men and women, but in some cases women tend to leave school earlier because of arranged marriages or the distances of school from their homes. Men predominantly work in agriculture and fishing, while women work in the kitchen. In NAB, even though women work on farmland they earn less then men. Weaving (NPS) and kitchen activity is recognized as women’s work, which also supports the household income. Mass media is not yielding any changes in TFS and food culture in these days. The traditional farming system has changed because of intensive agricultural extension which has introduced new methods of agriculture for the last three decades (since the 1980’s). There is no evidence that they want to change any of their food habits because of the mass media despite the lapau activity which allows them to get more food choices, instead preparing traditional meal at home. The recommendations of this thesis are: 1) The empowerment of farmers. It is regarding the self sufficient supply of manure, cooperative seed, and sustainable farm management. Farmers should know – where are they in their state of knowledge – so they can use their local wisdom and still collaborate with new sources of knowledge. Farmers should learn the prognosis of supply and demand next prior to harvest. There is a need for farm management guidelines; that can be adopted from both their local wisdom and modern knowledge. 2) Increase of non-agricultural income Increasing the non-agricultural income is strongly recommended. The remittances can be invested on non-agricultural jobs. 3) The empowerment of the mother. The mother plays an important role in farm to fork activities; the mother can be an initiator and promoter of cultivating spices in the backyard. Improvement of nutritional knowledge through information and informal public education can be done through arisan ibu-ibu and lapau activity. The challenges to apply these recommendations are: 1) The gap between institutions and organizations of local governments. There is more than one institution involved in food security policy. 2) Training and facilities for field extension agriculture (FEA) is needed because the rapid change of interaction between local government and farmer’s dependent on this agency.

From evolution to co-evolution – An empirical study of the role of innovation, policy and public research in industrial dynamics

In the past decades since Schumpeter’s influential writings economists have pursued research to examine the role of innovation in certain industries on firm as well as on industry level. Researchers describe innovations as the main trigger of industry dynamics, while policy makers argue that research and education are directly linked to economic growth and welfare. Thus, research and education are an important objective of public policy. Firms and public research are regarded as the main actors which are relevant for the creation of new knowledge. This knowledge is finally brought to the market through innovations. What is more, policy makers support innovations. Both actors, i.e. policy makers and researchers, agree that innovation plays a central role but researchers still neglect the role that public policy plays in the field of industrial dynamics. Therefore, the main objective of this work is to learn more about the interdependencies of innovation, policy and public research in industrial dynamics. The overarching research question of this dissertation asks whether it is possible to analyze patterns of industry evolution – from evolution to co-evolution – based on empirical studies of the role of innovation, policy and public research in industrial dynamics. This work starts with a hypothesis-based investigation of traditional approaches of industrial dynamics. Namely, the testing of a basic assumption of the core models of industrial dynamics and the analysis of the evolutionary patterns – though with an industry which is driven by public policy as example. Subsequently it moves to a more explorative approach, investigating co-evolutionary processes. The underlying questions of the research include the following: Do large firms have an advantage because of their size which is attributable to cost spreading? Do firms that plan to grow have more innovations? What role does public policy play for the evolutionary patterns of an industry? Are the same evolutionary patterns observable as those described in the ILC theories? And is it possible to observe regional co-evolutionary processes of science, innovation and industry evolution? Based on two different empirical contexts – namely the laser and the photovoltaic industry – this dissertation tries to answer these questions and combines an evolutionary approach with a co-evolutionary approach. The first chapter starts with an introduction of the topic and the fields this dissertation is based on. The second chapter provides a new test of the Cohen and Klepper (1996) model of cost spreading, which explains the relationship between innovation, firm size and R&D, at the example of the photovoltaic industry in Germany. First, it is analyzed whether the cost spreading mechanism serves as an explanation for size advantages in this industry. This is related to the assumption that the incentives to invest in R&D increase with the ex-ante output. Furthermore, it is investigated whether firms that plan to grow will have more innovative activities. The results indicate that cost spreading serves as an explanation for size advantages in this industry and, furthermore, growth plans lead to higher amount of innovative activities. What is more, the role public policy plays for industry evolution is not finally analyzed in the field of industrial dynamics. In the case of Germany, the introduction of demand inducing policy instruments stimulated market and industry growth. While this policy immediately accelerated market volume, the effect on industry evolution is more ambiguous. Thus, chapter three analyzes this relationship by considering a model of industry evolution, where demand-inducing policies will be discussed as a possible trigger of development. The findings suggest that these instruments can take the same effect as a technical advance to foster the growth of an industry and its shakeout. The fourth chapter explores the regional co-evolution of firm population size, private-sector patenting and public research in the empirical context of German laser research and manufacturing over more than 40 years from the emergence of the industry to the mid-2000s. The qualitative as well as quantitative evidence is suggestive of a co-evolutionary process of mutual interdependence rather than a unidirectional effect of public research on private-sector activities. Chapter five concludes with a summary, the contribution of this work as well as the implications and an outlook of further possible research.

Comentaris a la intervenció del professor Ken Bain: S'aprèn res estudiant a la universitat? = Comentarios a la intervención del profesor Ken Bain: Se aprende algo estudiando en la universidad? = Comments on professor Ken Bain's lecutre: Do we learn anything at university?

Comentaris a la videoconferència de Kenneth Bain sobre l'aprenentatge universitari

Adaptive intelligent agent approach to guide the web navigation on the PLAN-G distance learning platform

Hypermedia systems based on the Web for open distance education are becoming increasingly popular as tools for user-driven access learning information. Adaptive hypermedia is a new direction in research within the area of user-adaptive systems, to increase its functionality by making it personalized [Eklu 961. This paper sketches a general agents architecture to include navigational adaptability and user-friendly processes which would guide and accompany the student during hislher learning on the PLAN-G hypermedia system (New Generation Telematics Platform to Support Open and Distance Learning), with the aid of computer networks and specifically WWW technology [Marz 98-1] [Marz 98-2]. The PLAN-G actual prototype is successfully used with some informatics courses (the current version has no agents yet). The propased multi-agent system, contains two different types of adaptive autonomous software agents: Personal Digital Agents {Interface), to interacl directly with the student when necessary; and Information Agents (Intermediaries), to filtrate and discover information to learn and to adapt navigation space to a specific student

Introduction to Matlab - a self-study tutorial

Matlab is a high level language that is very easy to use and very powerful. It comes with a wealth of libraries and toolboxes, that you can use directly, so that you don't need to program low level functions. It enables you to display results very easily on graphs and images. To get started with it, you need to understand how to manipulate and represent data, and how to find information about the available functions. During this self-study tutorial, you will learn: 1- How to start Matlab. 2- How you can find out all the information you need. 3- How to create simple vectors and matrices. 4- What functions are available and how to find them. 5- How to plot graphs of functions. 6- How to write a script. After this (should take about an hour), you will know most of what you need to know about Matlab and should definitely know how to go on learning about it on your own…

How to Cite References - School of Geography

This is a quick guide for students to learn the appropriate Harvard referencing style in academic writing.

Example General Feedback on Routes to Success Portfolios

This document provides example feedback which has been generated following the marking of a class set of portfolios. It is used as a part of the Routes to Success Module, specifically on the section titled Sustaining Success. Students can read the feedback prior to completing the portfolio to alert them to the possible shortfalls which may occur when they undertake this type of task. The feedback is introduced in the context that the task of completing the portfolio is a developmental one, and that students can expect to learn and improve their performance for this type of task as they develop and refine their skills.

Tools to use when writing a thesis using Word 2011 (for Apple computers)

A collection of videos on time saving features when using MS Word 2011 to write a thesis. Learn how to use styles, make table of contents, make table of figures, use the document map, use the browse object tool and keep a count of the words in your file and many more useful features of Word 2011. Word 2011 is for Apple computers, there is a collection of similar video for use with the PC version Word 2010.