What is the Causal Effect of Information and Learning about a Public Good on Willingness to Pay?

In this study we elicit agents’ prior information set regarding a public good, exogenously give information treatments to survey respondents and subsequently elicit willingness to pay for the good and posterior information sets. The design of this field experiment allows us to perform theoretically motivated hypothesis testing between different updating rules: non-informative updating, Bayesian updating, and incomplete updating. We find causal evidence that agents imperfectly update their information sets. We also field causal evidence that the amount of additional information provided to subjects relative to their pre-existing information levels can affect stated WTP in ways consistent overload from too much learning. This result raises important (though familiar) issues for the use of stated preference methods in policy analysis.

Reducing the learning curve for the treatment of morphoeic (sclerosing) basal cell carcinoma of the face.

The treatment of morphoeic (or sclerosing) basal cell carcinoma (mBCC) of the face is associated with high rates of incomplete excision and recurrence. A principal risk factor for incomplete resection is the grade of surgeon. We did a prospective, randomised study of 40 consecutive patients with mBCC of the face. The extent of the tumour was assessed under standard conditions by consultant surgeons and compared with assessments by resident surgeons with the help of the Varioscope, a combination of microscope and loupe glasses with strong illumination and a maximal magnification of 7x. The data from a former retrospective study of all excisions of mBCC of the face during a five-year period at the hospital served as control. Residents with the support of the Varioscope achieved a rate of incomplete excisions similar to that of consultants under standard conditions. There was a significant reduction of the rate of incomplete resections by resident surgeons thanks to high magnification and good lighting (p=0.02). High magnification and good lighting were useful in learning how to recognise skin changes associated with mBCC of the face and achieving a low rate of incomplete excisions.

Extreme Precipitation Modelling Using Geostatistics and Machine Learning Algorithms

The paper presents an approach for mapping of precipitation data. The main goal is to perform spatial predictions and simulations of precipitation fields using geostatistical methods (ordinary kriging, kriging with external drift) as well as machine learning algorithms (neural networks). More practically, the objective is to reproduce simultaneously both the spatial patterns and the extreme values. This objective is best reached by models integrating geostatistics and machine learning algorithms. To demonstrate how such models work, two case studies have been considered: first, a 2-day accumulation of heavy precipitation and second, a 6-day accumulation of extreme orographic precipitation. The first example is used to compare the performance of two optimization algorithms (conjugate gradients and Levenberg-Marquardt) of a neural network for the reproduction of extreme values. Hybrid models, which combine geostatistical and machine learning algorithms, are also treated in this context. The second dataset is used to analyze the contribution of radar Doppler imagery when used as external drift or as input in the models (kriging with external drift and neural networks). Model assessment is carried out by comparing independent validation errors as well as analyzing data patterns.

Active Learning of Very-High Resolution Optical Imagery with SVM: Entropy vs Margin Sampling

An active learning method is proposed for the semi-automatic selection of training sets in remote sensing image classification. The method adds iteratively to the current training set the unlabeled pixels for which the prediction of an ensemble of classifiers based on bagged training sets show maximum entropy. This way, the algorithm selects the pixels that are the most uncertain and that will improve the model if added in the training set. The user is asked to label such pixels at each iteration. Experiments using support vector machines (SVM) on an 8 classes QuickBird image show the excellent performances of the methods, that equals accuracies of both a model trained with ten times more pixels and a model whose training set has been built using a state-of-the-art SVM specific active learning method

Towards Direct Policy Search Reinforcement Learning for Robot Control

This paper proposes a high-level reinforcement learning (RL) control system for solving the action selection problem of an autonomous robot. Although the dominant approach, when using RL, has been to apply value function based algorithms, the system here detailed is characterized by the use of direct policy search methods. Rather than approximating a value function, these methodologies approximate a policy using an independent function approximator with its own parameters, trying to maximize the future expected reward. The policy based algorithm presented in this paper is used for learning the internal state/action mapping of a behavior. In this preliminary work, we demonstrate its feasibility with simulated experiments using the underwater robot GARBI in a target reaching task

Aversive Stimuli and Functional Networks of Fear Learning in PTSD : 734

Background: One characteristic of post traumatic stress disorder is an inability to adapt to a safe environment i.e. to change behavior when predictions of adverse outcomes are not met. Recent studies have also indicated that PTSD patients have altered pain processing, with hyperactivation of the putamen and insula to aversive stimuli (Geuze et al, 2007). The present study examined neuronal responses to aversive and predicted aversive events. Methods: Twenty-four trauma exposed non-PTSD controls and nineteen subjects with PTSD underwent fMRI imaging during a partial reinforcement fear conditioning paradigm, with a mild electric shock as the unconditioned stimuli (UCS). Three conditions were analyzed: actual presentations of the UCS, events when a UCS was expected, but omitted (CS+), and events when the UCS was neither expected nor delivered (CS-). Results: The UCS evoked significant alterations in the pain matrix consisting of the brainstem, the midbrain, the thalamus, the insula, the anterior and middle cingulate and the contralateral somatosensory cortex. PTSD subjects displayed bilaterally elevated putamen activity to the electric shock, as compared to controls. In trials when USC was expected, but omitted, significant activations were observed in the brainstem, the midbrain, the anterior insula and the anterior cingulate. PTSD subjects displayed similar activations, but also elevated activations in the amygdala and the posterior insula. Conclusions: These results indicate altered fear and safety learning in PTSD, and neuronal activations are further explored in terms of functional connectivity using psychophysiological interaction analyses.

An empirical study of machine learning techniques applied to image watermarking

A study of how the machine learning technique, known as gentleboost, could improve different digital watermarking methods such as LSB, DWT, DCT2 and Histogram shifting.

Train-the-trainer: E-learning zur Förderung einer praxisorientiertem Vermittlung von evidenzbasierter Medizin

Hintergrund: Trotz ihrer Etablierung als essentieller Bestandteil der medizinischen Weiter-/Fortbildung werden europa- wie schweizweit kaum Kurse in evidenzbasierter Medizin (ebm) angeboten, die - integriert im klinischen Alltag - gezielt Fertigkeiten in ebm vermitteln. Noch grössere Defizite finden sich bei ebm- Weiterbildungsmöglichkeiten für klinische Ausbilder (z.B. Oberärzte). Als Weiterführung eines EU-finanzierten, klinisch integrierten E-learning- Programms für Weiterbildungsassistenten (www.ebm-unity.org) entwickelte eine europäische Gruppe von medical educators gezielt für Ausbilder ein e-learning-Curriculum zur Vermittlung von ebm im Rahmen der klinischen Weiterbildung. Methode: Die Entwicklung des Curriculums umfasst folgende Schritte: Beschreibung von Lernzielen, Identifikation von klinisch relevanten Lernumgebungen, Entwicklung von Lerninhalten und exemplarischen didaktischen Strategien, zugeschnitten auf die jeweilige Lernumgebungen, Design von web-basierten Selbst-Lernsequenzen mit Möglichkeiten zur Selbstevaluation, Erstellung eines Handbuchs. Ergebnisse: Lernziele des Tutoren-Lehrgangs sind der Erwerb von Fertigkeiten zur Vermittlung der 5 klassischen ebm-Schritte: PICO- (Patient-Intervention-Comparison-Outcome)-Fragen, Literatursuche, kritische Literaturbewertung, Übertragung der Ergebnisse im eigenen Setting und Implementierung). Die Lehrbeispiele zeigen angehenden ebm-Tutoren, wie sich typische klinische Situationen wie z.B. Stationsvisite, Ambulanzsprechstunde, Journalclub, offizielle Konferenzen, Audit oder das klinische Assessment von Weiterbildungsassistenten gezielt für die Vermittlung von ebm nutzen lassen. Kurze E-Learning-Module mit exemplarischen «real-life»-Video-Clips erlauben flexibles Lernen zugeschnitten auf das knappe Zeitkontingent von Ärzten. Eine Selbst-Evaluation ermöglicht die Überprüfung der gelernten Inhalte. Die Pilotierung des Tutoren-Lehrgangs mit klinisch tätigen Tutoren sowie die Übersetzung des Moduls in weitere Sprachen sind derzeit in Vorbereitung. chlussfolgerung: Der modulare Train-the-Trainer-Kurs zur Vermittlung von ebm im klinischen Alltag schliesst eine wichtige Lücke in der Dissemination von klinischer ebm. Webbasierte Beispiele mit kurzen Sequenzen demonstrieren typische Situationen zur Vermittlung der ebm-Kernfertigkeiten und bieten medical educators wie Oberärzten einen niedrigschwelligen Einstieg in «ebm» am Krankenbett. Langfristiges Ziel ist eine europäische Qualifikation für ebm- Learning und -Teaching in der Fort- und Weiterbildung. Nach Abschluss der Evaluation steht das Curriculum interessierten Personen und Gruppen unter «not-for-profit»-Bedingungen zur Verfügung. Auskünfte erhältlich von rkunz@uhbs.ch. Finanziert durch die Europäische Kommission - Leonardo da Vinci Programme - Transfer of Innovation - Pilot Project for Lifelong Learn- ing 2007 und das Schweizerische Staatssekretariat für Bildung und Forschung.

Kernel-based methods for change detection in remote sensing images

Nowadays, the joint exploitation of images acquired daily by remote sensing instruments and of images available from archives allows a detailed monitoring of the transitions occurring at the surface of the Earth. These modifications of the land cover generate spectral discrepancies that can be detected via the analysis of remote sensing images. Independently from the origin of the images and of type of surface change, a correct processing of such data implies the adoption of flexible, robust and possibly nonlinear method, to correctly account for the complex statistical relationships characterizing the pixels of the images. This Thesis deals with the development and the application of advanced statistical methods for multi-temporal optical remote sensing image processing tasks. Three different families of machine learning models have been explored and fundamental solutions for change detection problems are provided. In the first part, change detection with user supervision has been considered. In a first application, a nonlinear classifier has been applied with the intent of precisely delineating flooded regions from a pair of images. In a second case study, the spatial context of each pixel has been injected into another nonlinear classifier to obtain a precise mapping of new urban structures. In both cases, the user provides the classifier with examples of what he believes has changed or not. In the second part, a completely automatic and unsupervised method for precise binary detection of changes has been proposed. The technique allows a very accurate mapping without any user intervention, resulting particularly useful when readiness and reaction times of the system are a crucial constraint. In the third, the problem of statistical distributions shifting between acquisitions is studied. Two approaches to transform the couple of bi-temporal images and reduce their differences unrelated to changes in land cover are studied. The methods align the distributions of the images, so that the pixel-wise comparison could be carried out with higher accuracy. Furthermore, the second method can deal with images from different sensors, no matter the dimensionality of the data nor the spectral information content. This opens the doors to possible solutions for a crucial problem in the field: detecting changes when the images have been acquired by two different sensors.

Supporting the reuse of effective CSCL learning designs through social structure representations

Distance and blended collaborative learning settings are usually characterized by different social structures defined in terms of groups' number, dimension, and composition; these structures are variable and can change within the same activity. This variability poses additional complexity to instructional designers, when they are trying to develop successful experiences from existing designs. This complexity is greatly associated with the fact that learning designs do not render explicit how social structures influenced the decisions of the original designer, and thus whether the social structures of the new setting could preclude the effectiveness of the reused design. This article proposes the usage of new representations (social structure representations, SSRs) able to support unskilled designers in reusing existing learning designs, through the explicit characterization of the social structures and constraints embedded either by the original designers or the reusing teachers, according to well-known principles of good collaborative learning practice. The article also describes an evaluation process that involved university professors, as well as the main findings derived from it. This process supported the initial assumptions about the effectiveness of SSRs, with significant evidence from both qualitative and qualitative data.

LdShake: learning design solutions sharing and co-edition

Two important challenges that teachers are currently facing are the sharing and the collaborative authoring of their learning design solutions, such as didactical units and learning materials. On the one hand, there are tools that can be used for the creation of design solutions and only some of them facilitate the co-edition. However, they do not incorporate mechanisms that support the sharing of the designs between teachers. On the other hand, there are tools that serve as repositories of educational resources but they do not enable the authoring of the designs. In this paper we present LdShake, a web tool whose novelty is focused on the combined support for the social sharing and co-edition of learning design solutions within communities of teachers. Teachers can create and share learning designs with other teachers using different access rights so that they can read, comment or co-edit the designs. Therefore, each design solution is associated to a group of teachers able to work on its definition, and another group that can only see the design. The tool is generic in that it allows the creation of designs based on any pedagogical approach. However, it can be particularized in instances providing pre-formatted designs structured according to a specific didactic method (such as Problem-Based Learning, PBL). A particularized LdShake instance has been used in the context of Human Biology studies where teams of teachers are required to work together in the design of PBL solutions. A controlled user study, that compares the use of a generic LdShake and a Moodle system, configured to enable the creation and sharing of designs, has been also carried out. The combined results of the real and controlled studies show that the social structure, and the commenting, co-edition and publishing features of LdShake provide a useful, effective and usable approach for facilitating teachers' teamwork.

An Automatic Approach for Learning and Tuning Gaussian Interval Type-2 Fuzzy Membership Functions Applied to Lung CAD Classification System

The potential of type-2 fuzzy sets for managing high levels of uncertainty in the subjective knowledge of experts or of numerical information has focused on control and pattern classification systems in recent years. One of the main challenges in designing a type-2 fuzzy logic system is how to estimate the parameters of type-2 fuzzy membership function (T2MF) and the Footprint of Uncertainty (FOU) from imperfect and noisy datasets. This paper presents an automatic approach for learning and tuning Gaussian interval type-2 membership functions (IT2MFs) with application to multi-dimensional pattern classification problems. T2MFs and their FOUs are tuned according to the uncertainties in the training dataset by a combination of genetic algorithm (GA) and crossvalidation techniques. In our GA-based approach, the structure of the chromosome has fewer genes than other GA methods and chromosome initialization is more precise. The proposed approach addresses the application of the interval type-2 fuzzy logic system (IT2FLS) for the problem of nodule classification in a lung Computer Aided Detection (CAD) system. The designed IT2FLS is compared with its type-1 fuzzy logic system (T1FLS) counterpart. The results demonstrate that the IT2FLS outperforms the T1FLS by more than 30% in terms of classification accuracy.

Strong minimax lower bounds for learning

Minimax lower bounds for concept learning state, for example, thatfor each sample size $n$ and learning rule $g_n$, there exists a distributionof the observation $X$ and a concept $C$ to be learnt such that the expectederror of $g_n$ is at least a constant times $V/n$, where $V$ is the VC dimensionof the concept class. However, these bounds do not tell anything about therate of decrease of the error for a {\sl fixed} distribution--concept pair.\\In this paper we investigate minimax lower bounds in such a--stronger--sense.We show that for several natural $k$--parameter concept classes, includingthe class of linear halfspaces, the class of balls, the class of polyhedrawith a certain number of faces, and a class of neural networks, for any{\sl sequence} of learning rules $\{g_n\}$, there exists a fixed distributionof $X$ and a fixed concept $C$ such that the expected error is larger thana constant times $k/n$ for {\sl infinitely many n}. We also obtain suchstrong minimax lower bounds for the tail distribution of the probabilityof error, which extend the corresponding minimax lower bounds.

Adaptive learning in models with lagged variables

In this work I study the stability of the dynamics generated by adaptivelearning processes in intertemporal economies with lagged variables. Iprove that determinacy of the steady state is a necessary condition for the convergence of the learning dynamics and I show that the reciprocal is not true characterizing the economies where convergence holds. In the case of existence of cycles I show that there is not, in general, a relationship between determinacy and convergence of the learning process to the cycle. I also analyze the expectational stability of these equilibria.

Estimating learning models from experimental data

We study the statistical properties of three estimation methods for a model of learning that is often fitted to experimental data: quadratic deviation measures without unobserved heterogeneity, and maximum likelihood withand without unobserved heterogeneity. After discussing identification issues, we show that the estimators are consistent and provide their asymptotic distribution. Using Monte Carlo simulations, we show that ignoring unobserved heterogeneity can lead to seriously biased estimations in samples which have the typical length of actual experiments. Better small sample properties areobtained if unobserved heterogeneity is introduced. That is, rather than estimating the parameters for each individual, the individual parameters are considered random variables, and the distribution of those random variables is estimated.