Essais en microéconomie appliquée

La thèse comporte trois essais en microéconomie appliquée. En utilisant des modèles d’apprentissage (learning) et d’externalité de réseau, elle étudie le comportement des agents économiques dans différentes situations. Le premier essai de la thèse se penche sur la question de l’utilisation des ressources naturelles en situation d’incertitude et d’apprentissage (learning). Plusieurs auteurs ont abordé le sujet, mais ici, nous étudions un modèle d’apprentissage dans lequel les agents qui consomment la ressource ne formulent pas les mêmes croyances a priori. Le deuxième essai aborde le problème générique auquel fait face, par exemple, un fonds de recherche désirant choisir les meilleurs parmi plusieurs chercheurs de différentes générations et de différentes expériences. Le troisième essai étudie un modèle particulier d’organisation d’entreprise dénommé le marketing multiniveau (multi-level marketing). Le premier chapitre est intitulé "Renewable Resource Consumption in a Learning Environment with Heterogeneous beliefs". Nous y avons utilisé un modèle d’apprentissage avec croyances hétérogènes pour étudier l’exploitation d’une ressource naturelle en situation d’incertitude. Il faut distinguer ici deux types d’apprentissage : le adaptive learning et le learning proprement dit. Ces deux termes ont été empruntés à Koulovatianos et al (2009). Nous avons montré que, en comparaison avec le adaptive learning, le learning a un impact négatif sur la consommation totale par tous les exploitants de la ressource. Mais individuellement certains exploitants peuvent consommer plus la ressource en learning qu’en adaptive learning. En effet, en learning, les consommateurs font face à deux types d’incitations à ne pas consommer la ressource (et donc à investir) : l’incitation propre qui a toujours un effet négatif sur la consommation de la ressource et l’incitation hétérogène dont l’effet peut être positif ou négatif. L’effet global du learning sur la consommation individuelle dépend donc du signe et de l’ampleur de l’incitation hétérogène. Par ailleurs, en utilisant les variations absolues et relatives de la consommation suite à un changement des croyances, il ressort que les exploitants ont tendance à converger vers une décision commune. Le second chapitre est intitulé "A Perpetual Search for Talent across Overlapping Generations". Avec un modèle dynamique à générations imbriquées, nous avons étudié iv comment un Fonds de recherche devra procéder pour sélectionner les meilleurs chercheurs à financer. Les chercheurs n’ont pas la même "ancienneté" dans l’activité de recherche. Pour une décision optimale, le Fonds de recherche doit se baser à la fois sur l’ancienneté et les travaux passés des chercheurs ayant soumis une demande de subvention de recherche. Il doit être plus favorable aux jeunes chercheurs quant aux exigences à satisfaire pour être financé. Ce travail est également une contribution à l’analyse des Bandit Problems. Ici, au lieu de tenter de calculer un indice, nous proposons de classer et d’éliminer progressivement les chercheurs en les comparant deux à deux. Le troisième chapitre est intitulé "Paradox about the Multi-Level Marketing (MLM)". Depuis quelques décennies, on rencontre de plus en plus une forme particulière d’entreprises dans lesquelles le produit est commercialisé par le biais de distributeurs. Chaque distributeur peut vendre le produit et/ou recruter d’autres distributeurs pour l’entreprise. Il réalise des profits sur ses propres ventes et reçoit aussi des commissions sur la vente des distributeurs qu’il aura recrutés. Il s’agit du marketing multi-niveau (multi-level marketing, MLM). La structure de ces types d’entreprise est souvent qualifiée par certaines critiques de système pyramidal, d’escroquerie et donc insoutenable. Mais les promoteurs des marketing multi-niveau rejettent ces allégations en avançant que le but des MLMs est de vendre et non de recruter. Les gains et les règles de jeu sont tels que les distributeurs ont plus incitation à vendre le produit qu’à recruter. Toutefois, si cette argumentation des promoteurs de MLMs est valide, un paradoxe apparaît. Pourquoi un distributeur qui désire vraiment vendre le produit et réaliser un gain recruterait-il d’autres individus qui viendront opérer sur le même marché que lui? Comment comprendre le fait qu’un agent puisse recruter des personnes qui pourraient devenir ses concurrents, alors qu’il est déjà établi que tout entrepreneur évite et même combat la concurrence. C’est à ce type de question que s’intéresse ce chapitre. Pour expliquer ce paradoxe, nous avons utilisé la structure intrinsèque des organisations MLM. En réalité, pour être capable de bien vendre, le distributeur devra recruter. Les commissions perçues avec le recrutement donnent un pouvoir de vente en ce sens qu’elles permettent au recruteur d’être capable de proposer un prix compétitif pour le produit qu’il désire vendre. Par ailleurs, les MLMs ont une structure semblable à celle des multi-sided markets au sens de Rochet et Tirole (2003, 2006) et Weyl (2010). Le recrutement a un effet externe sur la vente et la vente a un effet externe sur le recrutement, et tout cela est géré par le promoteur de l’organisation. Ainsi, si le promoteur ne tient pas compte de ces externalités dans la fixation des différentes commissions, les agents peuvent se tourner plus ou moins vers le recrutement.

Single Frame Image Super Resolution Using Learned Directionlets

In this paper, a new directionally adaptive, learning based, single image super resolution method using multiple direction wavelet transform, called Directionlets is presented. This method uses directionlets to effectively capture directional features and to extract edge information along different directions of a set of available high resolution images .This information is used as the training set for super resolving a low resolution input image and the Directionlet coefficients at finer scales of its high-resolution image are learned locally from this training set and the inverse Directionlet transform recovers the super-resolved high resolution image. The simulation results showed that the proposed approach outperforms standard interpolation techniques like Cubic spline interpolation as well as standard Wavelet-based learning, both visually and in terms of the mean squared error (mse) values. This method gives good result with aliased images also.

An Electronic Market-Maker

This paper presents an adaptive learning model for market-making under the reinforcement learning framework. Reinforcement learning is a learning technique in which agents aim to maximize the long-term accumulated rewards. No knowledge of the market environment, such as the order arrival or price process, is assumed. Instead, the agent learns from real-time market experience and develops explicit market-making strategies, achieving multiple objectives including the maximizing of profits and minimization of the bid-ask spread. The simulation results show initial success in bringing learning techniques to building market-making algorithms.

The relationship between multi-species catch and effort: among fishery comparisons

The control of fishing mortality via fishing effort remains fundamental to most fisheries management strategies even at the local community or co-management level. Decisions to support such strategies require knowledge of the underlying response of the catch to changes in effort. Even under adaptive management strategies, imprecise knowledge of the response is likely to help accelerate the adaptive learning process. Data and institutional capacity requirements to employ multi-species biomass dynamics and age-structured models invariably render their use impractical particularly in less developed regions of the world. Surplus production models fitted to catch and effort data aggregated across all species offer viable alternatives. The current paper seeks models of this type that best describe the multi-species catch–effort responses in floodplain-rivers, lakes and reservoirs and reef-based fisheries based upon among fishery comparisons, building on earlier work. Three alternative surplus production models were fitted to estimates of catch per unit area (CPUA) and fisher density for 258 fisheries in Africa, Asia and South America. In all cases examined, the best or equal best fitting model was the Fox type, explaining up to 90% of the variation in CPUA. For lake and reservoir fisheries in Africa and Asia, the Schaefer and an asymptotic model fitted equally well. The Fox model estimates of fisher density (fishers km−2) at maximum yield (iMY) for floodplain-rivers, African lakes and reservoirs and reef-based fisheries are 13.7 (95% CI [11.8, 16.4]); 27.8 (95% CI [17.5, 66.7]) and 643 (95% CI [459,1075]), respectively and compare well with earlier estimates. Corresponding estimates of maximum yield are also given. The significantly higher value of iMY for reef-based fisheries compared to estimates for rivers and lakes reflects the use of a different measure of fisher density based upon human population size estimates. The models predict that maximum yield is achieved at a higher fishing intensity in Asian lakes compared to those in Africa. This may reflect the common practice in Asia of stocking lakes to augment natural recruitment. Because of the equilibrium assumptions underlying the models, all the estimates of maximum yield and corresponding levels of effort should be treated with caution.

Towards cost-sensitive adaptation: when is it worth updating your predictive model?

Our digital universe is rapidly expanding,more and more daily activities are digitally recorded, data arrives in streams, it needs to be analyzed in real time and may evolve over time. In the last decade many adaptive learning algorithms and prediction systems, which can automatically update themselves with the new incoming data, have been developed. The majority of those algorithms focus on improving the predictive performance and assume that model update is always desired as soon as possible and as frequently as possible. In this study we consider potential model update as an investment decision, which, as in the financial markets, should be taken only if a certain return on investment is expected. We introduce and motivate a new research problem for data streams ? cost-sensitive adaptation. We propose a reference framework for analyzing adaptation strategies in terms of costs and benefits. Our framework allows to characterize and decompose the costs of model updates, and to asses and interpret the gains in performance due to model adaptation for a given learning algorithm on a given prediction task. Our proof-of-concept experiment demonstrates how the framework can aid in analyzing and managing adaptation decisions in the chemical industry.

Zero attracting recursive least squares algorithms

The l1-norm sparsity constraint is a widely used technique for constructing sparse models. In this contribution, two zero-attracting recursive least squares algorithms, referred to as ZA-RLS-I and ZA-RLS-II, are derived by employing the l1-norm of parameter vector constraint to facilitate the model sparsity. In order to achieve a closed-form solution, the l1-norm of the parameter vector is approximated by an adaptively weighted l2-norm, in which the weighting factors are set as the inversion of the associated l1-norm of parameter estimates that are readily available in the adaptive learning environment. ZA-RLS-II is computationally more efficient than ZA-RLS-I by exploiting the known results from linear algebra as well as the sparsity of the system. The proposed algorithms are proven to converge, and adaptive sparse channel estimation is used to demonstrate the effectiveness of the proposed approach.

Neural network based glucose - insulin metabolism models for children with Type 1 diabetes

In this paper two models for the simulation of glucose-insulin metabolism of children with Type 1 diabetes are presented. The models are based on the combined use of Compartmental Models (CMs) and artificial Neural Networks (NNs). Data from children with Type 1 diabetes, stored in a database, have been used as input to the models. The data are taken from four children with Type 1 diabetes and contain information about glucose levels taken from continuous glucose monitoring system, insulin intake and food intake, along with corresponding time. The influences of taken insulin on plasma insulin concentration, as well as the effect of food intake on glucose input into the blood from the gut, are estimated from the CMs. The outputs of CMs, along with previous glucose measurements, are fed to a NN, which provides short-term prediction of glucose values. For comparative reasons two different NN architectures have been tested: a Feed-Forward NN (FFNN) trained with the back-propagation algorithm with adaptive learning rate and momentum, and a Recurrent NN (RNN), trained with the Real Time Recurrent Learning (RTRL) algorithm. The results indicate that the best prediction performance can be achieved by the use of RNN.

J-Quizmaker

Lehrvideos erfreuen sich dank aktueller Entwicklungen im Bereich der Online-Lehre (Videoplattformen, MOOCs) auf der einen Seite und einer riesigen Auswahl sowie einer einfachen Produktion und Distribution auf der anderen Seite großer Beliebtheit bei der Wissensvermittlung. Trotzdem bringen Videos einen entscheidenden Nachteil mit sich, welcher in der Natur des Datenformats liegt. So sind die Suche nach konkreten Sachverhalten in einem Video sowie die semantische Aufbereitung zur automatisierten Verknüpfung mit weiteren spezifischen Inhalten mit hohem Aufwand verbunden. Daher werden die lernerfolg-orientierte Selektion von Lehrsegmenten und ihr Arrangement zur auf Lernprozesse abgestimmten Steuerung gehemmt. Beim Betrachten des Videos werden unter Umständen bereits bekannte Sachverhalte wiederholt bzw. können nur durch aufwendiges manuelles Spulen übersprungen werden. Selbiges Problem besteht auch bei der gezielten Wiederholung von Videoabschnitten. Als Lösung dieses Problems wird eine Webapplikation vorgestellt, welche die semantische Aufbereitung von Videos hin zu adaptiven Lehrinhalten ermöglicht: mittels Integration von Selbsttestaufgaben mit definierten Folgeaktionen können auf Basis des aktuellen Nutzerwissens Videoabschnitte automatisiert übersprungen oder wiederholt und externe Inhalte verlinkt werden. Der präsentierte Ansatz basiert somit auf einer Erweiterung der behavioristischen Lerntheorie der Verzweigten Lehrprogramme nach Crowder, die auf den Lernverlauf angepasste Sequenzen von Lerneinheiten beinhaltet. Gleichzeitig werden mittels regelmäßig eingeschobener Selbsttestaufgaben Motivation sowie Aufmerksamkeit des Lernenden nach Regeln der Programmierten Unterweisung nach Skinner und Verstärkungstheorie gefördert. Durch explizite Auszeichnung zusammengehöriger Abschnitte in Videos können zusätzlich die enthaltenden Informationen maschinenlesbar gestaltet werden, sodass weitere Möglichkeiten zum Auffinden und Verknüpfen von Lerninhalten geschaffen werden.

A Computer-Based Strategy for Practical Skills Teaching

An adaptive learning technology embedded in e-learning environments ensures choice of the structure, content, and activities for each individual learner according to the teaching team’s domain and didactic knowledge and skills. In this paper a computer-based scenario for application of an adaptive navigation technology is proposed and demonstrated on an example course topic.

Education in Computer Science and the Role of the Teacher in the Environment of Open Educational Resources

Report published in the Proceedings of the National Conference on "Education in the Information Society", Plovdiv, May, 2013

Adaptability, Transformation and Complex Changes in Namibia and Tanzania: Resilience and Innovation System Development in Local Communities

This thesis is a research about the recent complex spatial changes in Namibia and Tanzania and local communities’ capacity to cope with, adapt to and transform the unpredictability engaged to these processes. I scrutinise the concept of resilience and its potential application to explaining the development of local communities in Southern Africa when facing various social, economic and environmental changes. My research is based on three distinct but overlapping research questions: what are the main spatial changes and their impact on the study areas in Namibia and Tanzania? What are the adaptation, transformation and resilience processes of the studied local communities in Namibia and Tanzania? How are innovation systems developed, and what is their impact on the resilience of the studied local communities in Namibia and Tanzania? I use four ethnographic case studies concerning environmental change, global tourism and innovation system development in Namibia and Tanzania, as well as mixed-methodological approaches, to study these issues. The results of my empirical investigation demonstrate that the spatial changes in the localities within Namibia and Tanzania are unique, loose assemblages, a result of the complex, multisided, relational and evolutional development of human and non-human elements that do not necessarily have linear causalities. Several changes co-exist and are interconnected though uncertain and unstructured and, together with the multiple stressors related to poverty, have made communities more vulnerable to different changes. The communities’ adaptation and transformation measures have been mostly reactive, based on contingency and post hoc learning. Despite various anticipation techniques, coping measures, adaptive learning and self-organisation processes occurring in the localities, the local communities are constrained by their uneven power relationships within the larger assemblages. Thus, communities’ own opportunities to increase their resilience are limited without changing the relations in these multiform entities. Therefore, larger cooperation models are needed, like an innovation system, based on the interactions of different actors to foster cooperation, which require collaboration among and input from a diverse set of stakeholders to combine different sources of knowledge, innovation and learning. Accordingly, both Namibia and Tanzania are developing an innovation system as their key policy to foster transformation towards knowledge-based societies. Finally, the development of an innovation system needs novel bottom-up approaches to increase the resilience of local communities and embed it into local communities. Therefore, innovation policies in Namibia have emphasised the role of indigenous knowledge, and Tanzania has established the Living Lab network.

Constant-random practice and the adaptive process in motor learning with varying amounts of constant practice

The adaptive process in motor learning was examined in terms of effects of varying amounts of constant practice performed before random practice. Participants pressed five response keys sequentially, the last one coincident with the lighting of a final visual stimulus provided by a complex coincident timing apparatus. Different visual stimulus speeds were used during the random practice. 33 children (M age=11.6 yr.) were randomly assigned to one of three experimental groups: constant-random, constant-random 33%, and constant-random 66%. The constant-random group practiced constantly until they reached a criterion of performance stabilization three consecutive trials within 50 msec. of error. The other two groups had additional constant practice of 33 and 66%, respectively, of the number of trials needed to achieve the stabilization criterion. All three groups performed 36 trials under random practice; in the adaptation phase, they practiced at a different visual stimulus speed adopted in the stabilization phase. Global performance measures were absolute, constant, and variable errors, and movement pattern was analyzed by relative timing and overall movement time. There was no group difference in relation to global performance measures and overall movement time. However, differences between the groups were observed on movement pattern, since constant-random 66% group changed its relative timing performance in the adaptation phase.

Cost dependent strategy for electricity markets bidding based on adaptive reinforcement learning

Electricity markets are complex environments, involving a large number of different entities, playing in a dynamic scene to obtain the best advantages and profits. MASCEM is a multi-agent electricity market simulator to model market players and simulate their operation in the market. Market players are entities with specific characteristics and objectives, making their decisions and interacting with other players. MASCEM provides several dynamic strategies for agents’ behavior. This paper presents a method that aims to provide market players with strategic bidding capabilities, allowing them to obtain the higher possible gains out of the market. This method uses a reinforcement learning algorithm to learn from experience how to choose the best from a set of possible bids. These bids are defined accordingly to the cost function that each producer presents.

Learning objects recommendation in an adaptive educational hypermedia system

The aim of this paper is presenting the modules of the Adaptive Educational Hypermedia System PCMAT, responsible for the recommendation of learning objects. PCMAT is an online collaborative learning platform with a constructivist approach, which assesses the user’s knowledge and presents contents and activities adapted to the characteristics and learning style of students of mathematics in basic schools. The recommendation module and search and retrieval module choose the most adequate learning object, based on the user's characteristics and performance, and in this way contribute to the system’s adaptability.