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.

Cross-curricular approaches to teaching and learning.

Para profesores de primaria y todos aquellos involucrados en el diseño curricular. Aporta ideas prácticas sobre cómo el maestro puede utilizar la historia y la geografía como materias transversales para proporcionar un vehículo a través del cual los niños pueden aplicar los conocimientos y conceptos adquiridos en estas áreas. Además, los niños toman conciencia de cómo utilizar, desarrollar y ampliar los conocimientos que están adquiriendo de modo que el aprendizaje puede ser más integrado y relevante para ellos.

Inclusive approaches to teaching and learning mathematics

This chapter explores the role of mentors in supporting pre-service teachers to include all children in mathematics teaching, no matter what their individual needs.

Be careful how you ask! Using focus groups and nominal group technique to explore the barriers to learning

Schools have a legal duty to make reasonable adjustments for disabled pupils who experience barriers to learning. Inclusive approaches to data collection ensure that the needs of all children who are struggling are not overlooked. However, it is important that the methods promote sustained reflection on the part of all children, do not inadvertently accentuate differences between pupils, and do not allow individual needs to go unrecognized. This paper examines more closely the processes involved in using Nominal Group Technique to collect the views of children with and without a disability on the difficulties experienced in school. Data were collected on the process as well as the outcomes of using this technique to examine how pupil views are transformed from the individual to the collective, a process that involves making the private, public. Contrasts are drawn with questionnaire data, another method of data collection favoured by teachers. Although more time-efficient this can produce unclear and cursory responses. The views that surface from pupils need also to be seen within the context of the ways in which schools customize the data collection process and the ways in which the format and organization of the activity impact on the responses and responsiveness of the pupils.

Developing critical understanding in HRM students:using innovative teaching methods to encourage deep approaches to study

Purpose – This paper aims to focus on developing critical understanding in human resource management (HRM) students in Aston Business School, UK. The paper reveals that innovative teaching methods encourage deep approaches to study, an indicator of students reaching their own understanding of material and ideas. This improves student employability and satisfies employer need. Design/methodology/approach – Student response to two second year business modules, matched for high student approval rating, was collected through focus group discussion. One module was taught using EBL and the story method, whilst the other used traditional teaching methods. Transcripts were analysed and compared using the structure of the ASSIST measure. Findings – Critical understanding and transformative learning can be developed through the innovative teaching methods of enquiry-based learning (EBL) and the story method. Research limitations/implications – The limitation is that this is a single case study comparing and contrasting two business modules. The implication is that the study should be replicated and developed in different learning settings, so that there are multiple data sets to confirm the research finding. Practical implications – Future curriculum development, especially in terms of HE, still needs to encourage students and lecturers to understand more about the nature of knowledge and how to learn. The application of EBL and the story method is described in a module case study – “Strategy for Future Leaders”. Originality/value – This is a systematic and comparative study to improve understanding of how students and lecturers learn and of the context in which the learning takes place.

Alternative approaches to education: Tolstoy’s thinking on teaching and learning and its relevance for today

In this article I reflect upon the educational writings and teaching experiences of the 19th-Century Russian novelist Leo Tolstoy. Tolstoy is known to have attached much importance to his own writing on education, even more than to the literary creations for which he is best remembered. His writings on education have much to contribute to our present-day understanding of the learning process and cover such issues as, ‘learner autonomy’, ‘motivation’, ‘relationship’ and ‘student voice’. Tolstoy’s teaching experience was with multiethnic peasant children in his schools in Yasnaya Polyana. I intend to illustrate that the themes and issues that arose from his experiences in the 1860s can still find resonance with students and teachers in the 21st century.

Deep learning and spatial transcriptomics to investigate thyroid tumors

There are many diseases that affect the thyroid gland, and among them are carcinoma. Thyroid cancer is the most common endocrine neoplasm and the second most frequent cancer in the 0-49 age group. This thesis deals with two studies I conducted during my PhD. The first concerns the development of a Deep Learning model to be able to assist the pathologist in screening of thyroid cytology smears. This tool created in collaboration with Prof. Diciotti, affiliated with the DEI-UNIBO "Guglielmo Marconi" Department of Electrical Energy and Information Engineering, has an important clinical implication in that it allows patients to be stratified between those who should undergo surgery and those who should not. The second concerns the application of spatial transcriptomics on well-differentiated thyroid carcinomas to better understand their invasion mechanisms and thus to better comprehend which genes may be involved in the proliferation of these tumors. This project specifically was made possible through a fruitful collaboration with the Gustave Roussy Institute in Paris. Studying thyroid carcinoma deeply is essential to improve patient care, increase survival rates, and enhance the overall understanding of this prevalent cancer. It can lead to more effective prevention, early detection, and treatment strategies that benefit both patients and the healthcare system.

New approaches to underground systems in Brazilian Smilax species (Smilacaceae)

MARTINS, A. R. (Institute of Biology, State University of Campinas - UNICAMP, 13083-970, Campinas, SP, Brazil), N. PUT, (Division of Biology and Education, University of Vechta, 49377 Vechta, Germany), A. N. SOARES, A.B BOMB, and B. APPEZZATO DA GLORIA (Biological Science Department, Escola Superior de Agricultura `Luiz de Queiroz`, University of Sao Paulo, 13418-900, Piracicaba, SP, Brazil). J. Torrey Bot. Soc. 137: 220-235. 2010.-New approaches to underground systems in Brazilian Smilax species (Smilacaceae). Scientific studies show that the watery extract of the thickened underground stem and its adventitious roots of the genus Smilax can act as a therapeutic agent in immunoinflammatory disorders, such as rheumatic arthritis. Brazilians have used this genus of plants in folk medicine, however it is very hard to identify these species, since the morphology of the underground systems is very similar in this group. For better identification of those systems, we studied six species of Smilax L. (S. brasiliensis, S. campestris, S. cissoides, S. goyazana, S. oblongifolia and S. rufescens), collected in different regions of Brazil with different physiognomies and soil characteristics. The main purpose is to describe the morpho-anatomy of the underground systems and to analyze if their structure depends on environmental conditions. The underground stem (rhizophore) is of brown color and it is knotty, massive, slender (S. rufescens) or tuberous (S. brasiliensis, S. campestris, S. cissoides, S. goyazana and S. oblongifolia). The tuberization is a result of primary thickened meristem (PTM) activity. The color and thickness of the adventitious roots change during development because the epidermis and outer cortex are disposed of, so the inner cortex becomes the new covering tissue with lignified and dark color cells. There are differences in starch grain shapes in mature roots. The chemical attributes of the soil are very similar in all studied environments and, even when soil characteristics varied, all the species` underground system was distributed close to the soil surface (10 to 15 cm deep). The species exhibited clonal growth hence their underground system functions as storage structures and the axillary buds can sprout into new stems. Only Smilax rufescens, collected in sandy soil of Restinga, has vegetative dispersal due to the runners.

Immunity to asexual blood stage malaria and vaccine approaches

The development of a malaria vaccine seems to be a definite possibility despite the fact that even individuals with a life time of endemic exposure do not develop sterile immunity. An effective malaria vaccine would be invaluable in preventing malaria-associated deaths in endemic areas, especially amongst children less than 5 years of age and pregnant women. This review discusses our current understanding of immunity against the asexual blood stage of malaria - the stage that is responsible for the symptoms of the disease - and approaches to the design of an asexual blood stage vaccine.

Digital serious games and simulation games - comparison of two approaches to lean training

In an increasingly competitive and globalized world, companies need effective training methodologies and tools for their employees. However, selecting the most suitable ones is not an easy task. It depends on the requirements of the target group (namely time restrictions), on the specificities of the contents, etc. This is typically the case for training in Lean, the waste elimination manufacturing philosophy. This paper presents and compares two different approaches to lean training methodologies and tools: a simulation game based on a single realistic manufacturing platform, involving production and assembly operations that allows learning by playing; and a digital game that helps understand lean tools. This paper shows that both tools have advantages in terms of trainee motivation and knowledge acquisition. Furthermore, they can be used in a complementary way, reinforcing the acquired knowledge.

Modern immunological approaches to assess malaria transmission and immunity and to diagnose plasmodial infection

The present paper reviews our recent data concerning the use of immunological methods employing monoclonal antibodies and synthetic peptides to study malaria transmission and immunity and to diagnose plasmodial infection. As concerns malaria transmission, we studied the main vectors of human malaria and the plasmodial species transmitted in endemic areas of Rondônia state, Brazil. The natural infection on anopheline was evaluated by immunoradiometric assay (IRMA) using monoclonal antibodies to an immunodominant sporozoite surface antigen (CS protein) demonstrated to be species specific. Our results showed that among six species of Anopheles found infected, An. darlingi was the main vector transmitting Plasmodium falciparum and P. vivax malaria in the immediate vicinity of houses. In order to assess the level of anti-CS antibodies we studied, by IRMA using the synthetic peptide corresponding to the repetitive epitope of the sporozoite CS protein, sera of individuals living in the same areas where the entomological survey has been performed. In this assay the prevalence of anti-CS antibodies was very low and did not reflect the malaria transmission rate in the studied areas. In relation to malaria diagnosis, a monoclonal antibody specific to an epitope of a 50 kDa exoantigen, the major component of supernatant collected at the time of schizont rupture, was used as a probe for the detection of P. falciparum antigens. This assay seemed to be more sensitive than parasitological examination for malaria diagnosis since it was able to detect plasmodial antigens in both symptomatic and asymtomatic individuals with negative thick blood smear at different intervals after a last parasitologically confirmed confirmed attack of malaria.

From deep seated slope deformation to rock avalanche: Destabilization and transportation models of the Sierre landslide (Switzerland)

Sackung is a widespread post-glacial morphological feature affecting Alpine mountains and creating characteristic geomorphological expression that can be detected from topography. Over long time evolution, internal deformation can lead to the formation of rapidly moving phenomena such as a rock-slide or rock avalanche. In this study, a detailed description of the Sierre rock-avalanche (SW Switzerland) is presented. This convex-shaped postglacial instability is one of the larger rock-avalanche in the Alps, involving more than 1.5 billion m3 with a run-out distance of about 14 km and extremely low Fahrböschung angle. This study presents comprehensive analyses of the structural and geological characteristics leading to the development of the Sierre rock-avalanche. In particular, by combining field observations, digital elevation model analyses and numerical modelling, the strong influence of both ductile and brittle tectonic structures on the failure mechanism and on the failure surface geometry is highlighted. The detection of pre-failure deformation indicates that the development of the rock avalanche corresponds to the last evolutionary stage of a pre-existing deep seated gravitational slope instability. These analyses accompanied by the dating and the characterization of rock avalanche deposits, allow the proposal of a destabilization model that clarifies the different phases leading to the development of the Sierre rock avalanche.

Effects of web-based electrocardiography simulation on strategies and learning styles

OBJECTIVETo identify the association between the use of web simulation electrocardiography and the learning approaches, strategies and styles of nursing degree students.METHODA descriptive and correlational design with a one-group pretest-posttest measurement was used. The study sample included 246 students in a Basic and Advanced Cardiac Life Support nursing class of nursing degree.RESULTSNo significant differences between genders were found in any dimension of learning styles and approaches to learning. After the introduction of web simulation electrocardiography, significant differences were found in some item scores of learning styles: theorist (p < 0.040), pragmatic (p < 0.010) and approaches to learning.CONCLUSIONThe use of a web electrocardiogram (ECG) simulation is associated with the development of active and reflexive learning styles, improving motivation and a deep approach in nursing students.

Application of Multi-SNP Approaches Bayesian LASSO and AUC-RF to Detect Main Effects of Inflammatory-Gene Variants Associated with Bladder Cancer Risk

The relationship between inflammation and cancer is well established in several tumor types, including bladder cancer. We performed an association study between 886 inflammatory-gene variants and bladder cancer risk in 1,047 cases and 988 controls from the Spanish Bladder Cancer (SBC)/EPICURO Study. A preliminary exploration with the widely used univariate logistic regression approach did not identify any significant SNP after correcting for multiple testing. We further applied two more comprehensive methods to capture the complexity of bladder cancer genetic susceptibility: Bayesian Threshold LASSO (BTL), a regularized regression method, and AUC-Random Forest, a machine-learning algorithm. Both approaches explore the joint effect of markers. BTL analysis identified a signature of 37 SNPs in 34 genes showing an association with bladder cancer. AUC-RF detected an optimal predictive subset of 56 SNPs. 13 SNPs were identified by both methods in the total population. Using resources from the Texas Bladder Cancer study we were able to replicate 30% of the SNPs assessed. The associations between inflammatory SNPs and bladder cancer were reexamined among non-smokers to eliminate the effect of tobacco, one of the strongest and most prevalent environmental risk factor for this tumor. A 9 SNP-signature was detected by BTL. Here we report, for the first time, a set of SNP in inflammatory genes jointly associated with bladder cancer risk. These results highlight the importance of the complex structure of genetic susceptibility associated with cancer risk.

Neurotrophic activity of human adipose stem cells isolated from deep and superficial layers of abdominal fat.

New approaches to the clinical treatment of traumatic nerve injuries may one day utilize stem cells to enhance nerve regeneration. Adipose-derived stem cells (ASC) are found in abundant quantities and can be harvested by minimally invasive procedures that should facilitate their use in such regenerative applications. We have analyzed the properties of human ASC isolated from the deep and superficial layers of abdominal fat tissue obtained during abdominoplasty procedures. Cells from the superficial layer proliferate significantly faster than those from the deep layer. In both the deep and superficial layers, ASC express the pluripotent stem cell markers oct4 and nanog and also the stro-1 cell surface antigen. Superficial layer ASC induce the significantly enhanced outgrowth of neurite-like processes from neuronal cell lines when compared with that of deep layer cells. However, analysis by reverse transcription with the polymerase chain reaction and by enzyme-linked immunosorbent assay has revealed that ASC isolated from both layers express similar levels of the following neurotrophic factors: nerve growth factor, brain-derived neurotrophic factor and glial-derived neurotrophic factor. Thus, human ASC show promising potential for the treatment of traumatic nerve injuries. In particular, superficial layer ASC warrant further analysis of their neurotrophic molecules.