Neural substrates of social emotion regulation: a FMRI study on imitation and expressive suppression to dynamic facial signals.

Emotion regulation is crucial for successfully engaging in social interactions. Yet, little is known about the neural mechanisms controlling behavioral responses to emotional expressions perceived in the face of other people, which constitute a key element of interpersonal communication. Here, we investigated brain systems involved in social emotion perception and regulation, using functional magnetic resonance imaging (fMRI) in 20 healthy participants. The latter saw dynamic facial expressions of either happiness or sadness, and were asked to either imitate the expression or to suppress any expression on their own face (in addition to a gender judgment control task). fMRI results revealed higher activity in regions associated with emotion (e.g., the insula), motor function (e.g., motor cortex), and theory of mind (e.g., [pre]cuneus) during imitation. Activity in dorsal cingulate cortex was also increased during imitation, possibly reflecting greater action monitoring or conflict with own feeling states. In addition, premotor regions were more strongly activated during both imitation and suppression, suggesting a recruitment of motor control for both the production and inhibition of emotion expressions. Expressive suppression (eSUP) produced increases in dorsolateral and lateral prefrontal cortex typically related to cognitive control. These results suggest that voluntary imitation and eSUP modulate brain responses to emotional signals perceived from faces, by up- and down-regulating activity in distributed subcortical and cortical networks that are particularly involved in emotion, action monitoring, and cognitive control.

Emergence of oriented circuits driven by synaptic pruning associated with spike-timing-dependent plasticity (STDP)

ABSTRACT: Massive synaptic pruning following over-growth is a general feature of mammalian brain maturation. Pruning starts near time of birth and is completed by time of sexual maturation. Trigger signals able to induce synaptic pruning could be related to dynamic functions that depend on the timing of action potentials. Spike-timing-dependent synaptic plasticity (STDP) is a change in the synaptic strength based on the ordering of pre- and postsynaptic spikes. The relation between synaptic efficacy and synaptic pruning suggests that the weak synapses may be modified and removed through competitive "learning" rules. This plasticity rule might produce the strengthening of the connections among neurons that belong to cell assemblies characterized by recurrent patterns of firing. Conversely, the connections that are not recurrently activated might decrease in efficiency and eventually be eliminated. The main goal of our study is to determine whether or not, and under which conditions, such cell assemblies may emerge out of a locally connected random network of integrate-and-fire units distributed on a 2D lattice receiving background noise and content-related input organized in both temporal and spatial dimensions. The originality of our study stands on the relatively large size of the network, 10,000 units, the duration of the experiment, 10E6 time units (one time unit corresponding to the duration of a spike), and the application of an original bio-inspired STDP modification rule compatible with hardware implementation. A first batch of experiments was performed to test that the randomly generated connectivity and the STDP-driven pruning did not show any spurious bias in absence of stimulation. Among other things, a scale factor was approximated to compensate for the network size on the ac¬tivity. Networks were then stimulated with the spatiotemporal patterns. The analysis of the connections remaining at the end of the simulations, as well as the analysis of the time series resulting from the interconnected units activity, suggest that feed-forward circuits emerge from the initially randomly connected networks by pruning. RESUME: L'élagage massif des synapses après une croissance excessive est une phase normale de la ma¬turation du cerveau des mammifères. L'élagage commence peu avant la naissance et est complété avant l'âge de la maturité sexuelle. Les facteurs déclenchants capables d'induire l'élagage des synapses pourraient être liés à des processus dynamiques qui dépendent de la temporalité rela¬tive des potentiels d'actions. La plasticité synaptique à modulation temporelle relative (STDP) correspond à un changement de la force synaptique basé sur l'ordre des décharges pré- et post- synaptiques. La relation entre l'efficacité synaptique et l'élagage des synapses suggère que les synapses les plus faibles pourraient être modifiées et retirées au moyen d'une règle "d'appren¬tissage" faisant intervenir une compétition. Cette règle de plasticité pourrait produire le ren¬forcement des connexions parmi les neurones qui appartiennent à une assemblée de cellules caractérisée par des motifs de décharge récurrents. A l'inverse, les connexions qui ne sont pas activées de façon récurrente pourraient voir leur efficacité diminuée et être finalement éliminées. Le but principal de notre travail est de déterminer s'il serait possible, et dans quelles conditions, que de telles assemblées de cellules émergent d'un réseau d'unités integrate-and¬-fire connectées aléatoirement et distribuées à la surface d'une grille bidimensionnelle recevant à la fois du bruit et des entrées organisées dans les dimensions temporelle et spatiale. L'originalité de notre étude tient dans la taille relativement grande du réseau, 10'000 unités, dans la durée des simulations, 1 million d'unités de temps (une unité de temps correspondant à une milliseconde), et dans l'utilisation d'une règle STDP originale compatible avec une implémentation matérielle. Une première série d'expériences a été effectuée pour tester que la connectivité produite aléatoirement et que l'élagage dirigé par STDP ne produisaient pas de biais en absence de stimu¬lation extérieure. Entre autres choses, un facteur d'échelle a pu être approximé pour compenser l'effet de la variation de la taille du réseau sur son activité. Les réseaux ont ensuite été stimulés avec des motifs spatiotemporels. L'analyse des connexions se maintenant à la fin des simulations, ainsi que l'analyse des séries temporelles résultantes de l'activité des neurones, suggèrent que des circuits feed-forward émergent par l'élagage des réseaux initialement connectés au hasard.

Biological Flora of the British Isles: Ambrosia artemisiifolia

1. This account presents information on all aspects of the biology of Ambrosia artemisiifolia L. (Common ragweed) that are relevant to understanding its ecology. The main topics are presented within the standard framework of the Biological Flora of the British Isles: distribution, habitat, communities, responses to biotic factors, responses to environment, structure and physiology, phenology, floral and seed characters, herbivores and disease, history, and conservation, impacts and management. 2. Ambrosia artemisiifolia is a monoecious, wind-pollinated, annual herb native to North America whose height varies from 10 cm to 2.5 m according to environmental conditions. It has erect, branched stems and pinnately lobed leaves. Spike-like racemes of male capitula composed of staminate (male) florets terminate the stems, while cyme-like clusters of pistillate (female) florets are arranged in groups the axils of main and lateral stem leaves. 3. Seeds require prolonged chilling to break dormancy. Following seedling emergence in spring, the rate of vegetative growth depends on temperature, but development occurs over a wide thermal range. In temperate European climates, male and female flowers are produced from summer to early autumn (July to October). 4. Ambrosia artemisiifolia is sensitive to freezing. Late spring frosts kill seedlings and the first autumn frosts terminate the growing season. It has a preference for dry soils of intermediate to rich nutrient level. 5. Ambrosia artemisiifolia was introduced into Europe with seed imports from North America in the 19th century. Since World War II, it has become widespread in temperate regions of Europe and is now abundant in open, disturbed habitats as a ruderal and agricultural weed. 6. Recently, the N. American ragweed leaf beetle (Ophraella communa) has been detected in southern Switzerland and northern Italy. This species appears to have the capacity to substantially reduce growth and seed production of A. artemisiifolia. 7. In heavily infested regions of Europe, A. artemisiifolia causes substantial crop-yield losses and its copious, highly allergenic pollen creates considerable public health problems. There is consensus among models that climate change will allow its northward and up-hill spread in Europe.

Morphostructure and paleobiology of Mesogean orthophragminids (Discocyclinidae and Orbitoclypeidae, Foraminifera)

Orthophragminids are Paleocene-Eocene bilamellarperforate larger foraminifers characterized by a lenticular test with an equatorial layer of rectangular chamberlets arranged in concentric rings and lateral chamberlets on either side. Usually called 'discocyclinids', orthophragrninids include genera from two families: Discocyclinidae and Orbitoclypeidae.

Lateral gene exchanges shape the genomes of amoeba-resisting microorganisms.

Based on Darwin's concept of the tree of life, vertical inheritance was thought to be dominant, and mutations, deletions, and duplication were streaming the genomes of living organisms. In the current genomic era, increasing data indicated that both vertical and lateral gene inheritance interact in space and time to trigger genome evolution, particularly among microorganisms sharing a given ecological niche. As a paradigm to their diversity and their survival in a variety of cell types, intracellular microorganisms, and notably intracellular bacteria, were considered as less prone to lateral genetic exchanges. Such specialized microorganisms generally have a smaller gene repertoire because they do rely on their host's factors for some basic regulatory and metabolic functions. Here we review events of lateral gene transfer (LGT) that illustrate the genetic exchanges among intra-amoebal microorganisms or between the microorganism and its amoebal host. We tentatively investigate the functions of laterally transferred genes in the light of the interaction with their host as they should confer a selective advantage and success to the amoeba-resisting microorganisms (ARMs).

Apolipoprotein E and Recovery from Traumatic Brain Injury

The outcome from traumatic brain injury (TBI) is variable and only partly explained by known prognostic factors. This is especially true for predicting long-term outcome. Genetic factors may influence the brain`s susceptibility to injury or capacity for repair and regeneration. To examine the association of apolipoproteinE (apoE) genotype with long-term outcome, hippocampal volumes and general brain atrophy, we determined the apoE genotype from 61 TBI patients who had been injured over on average 31 years earlier. The long-term outcome was evaluated with repeated neuropsychological testing and by applying various measures of everyday functioning and quality of life. Magnetic resonance imaging (MRI) based volumetric analyses of the hippocampus and lateral ventricles were performed. In the prospective study, the purpose was to examine the association between apoE genotype and visibility of traumatic brain lesions during the first year after TBI and the ability of apoE genotype, the Glasgow Coma Score (GCS), MRI findings and duration of posttraumatic amnesia (PTA) to predict the one-year outcome. Thirty-three patients with TBI were studied and the outcome was evaluated with the Head Injury Symptom Checklist (HISC) and the Glasgow Outcome Scale extended version (GOS-E) scores one year after the injury. MRI and apoE genotyping were carried out. After three decades, neither hippocampal nor lateral ventricle volumes differed significantly in those patients with the apoE ε4 allele vs those without this allele, but the TBI patients with the apoE ε4 allele showed significantly poorer general cognitive level than those without this allele. This decline was wholly accounted for by a subgroup of patients who had developed incident or clinical dementia. In the prospective study the apoE genotype was not associated with visible MRI changes or outcome. The duration of PTA and acute MRI were the best predictors of one-year outcome in TBI. A portion of the TBI patients with the apoE ε4 allele seem to be at risk of long-term cognitive decline. This association may involve mechanisms other than those responsible for the development of brain atrophy. The early MRI and PTA have an important role in assessing the injury severity and prognosis.

Mild Cognitive Impairment and Early Detection of Alzheimer`s Disease. A Positron Emission Tomography Study

Alzheimer`s disease (AD) is characterised neuropathologically by the presence of extracellular amyloid plaques, intraneuronal neurofibrillary tangles, and cerebral neuronal loss. The pathological changes in AD are believed to start even decades before clinical symptoms are detectable. AD gradually affects episodic memory, cognition, behaviour and the ability to perform everyday activities. Mild cognitive impairment (MCI) represents a transitional state between normal aging and dementia disorders, especially AD. The predictive accuracy of the current and commonly used MCI criteria devide this disorder into amnestic (aMCI) and non-amnestic (naMCI) MCI. It seems that many individuals with aMCI tend to convert to AD. However many MCI individuals will remain stable and some may even recover. At present, the principal drugs for the treatment of AD provide only symptomatic and palliative benefits. Safe and effective mechanism-based therapies are needed for this devastating neurodegenerative disease of later life. In conjunction with the development of new therapeutic drugs, tools for early detection of AD would be important. In future one of the challenges will be to detect at an early stage these MCI individuals who will convert to AD. Methods which can predict which MCI subjects will convert to AD will be much more important if the new drug candidates prove to have disease-arresting or even disease–slowing effects. These types of drugs are likely to have the best efficacy if administered in the early or even in the presymptomatic phase of the disease when the synaptic and neuronal loss has not become too widespread. There is no clinical method to determine with certainly which MCI individuals will progress to AD. However there are several methods which have been suggested as predictors of conversion to AD, e.g. increased [11C] PIB uptake, hippocampal atrophy in MRI, low CSF A beta 42 level, high CSF tau-protein level, apolipoprotein E (APOE) ε4 allele and impairment in episodic memory and executive functions. In the present study subjects with MCI appear to have significantly higher [¹¹C] PIB uptake vs healthy elderly in several brain areas including frontal cortex, the posterior cingulate, the parietal and lateral temporal cortices, putamen and caudate. Also results from this PET study indicate that over time, MCI subjects who display increased [¹¹C] PIB uptake appear to be significantly more likely to convert to AD than MCI subjects with negative [¹¹C] PIB retention. Also hippocampal atrophy seems to increase in MCI individuals clearly during the conversion to AD. In this study [¹¹C] PIB uptake increases early and changes relatively little during the AD process whereas there is progressive hippocampal atrophy during the disease. In addition to increased [¹¹C] PIB retention and hippocampal atrophy, the status of APOE ε4 allele might contribute to the conversion from MCI to AD.

DYNAMICS OF LEAF FALL FROM RIPARIAN VEGETATION AND THE ACCUMULATION IN BENTHIC STOCK IN NEOTROPICAL STREAMS1

ABSTRACT Considering the importance of the riparian vegetation leaves as an energetic source to first order streams, the aim of the present study was to evaluate the leaf biomass contribution to the system and its temporal dynamics. With this purpose, monthly samples from July 2008 to June 2009 were collected using four sampling devices installed in three streams, in order to collect the vertical, lateral and terrestrial loads, and the benthic stock. We tested the following hypothesis: (1) leaf biomass input is higher after hydric stress periods; and (2) benthic stock biomass increase with higher loads from vertical and lateral entrances. Leaves represented 71.9% (on average) of all sampled allochthonous matter, with seasonal significant variation along the studied year. Peaks of leaf input were registered in September-October, after an increase in rainfall, and also in January, after a decrease in rainfall. Leaf input was higher in the lateral load.

Accidental and experimental Closantel intoxication in Uruguayan sheep

Abstract An outbreak of Closantel intoxication in sheep in Uruguay is described. The outbreak occurred in a group of 1300 weaning lambs treated orally with a 10% solution of Closantel. One hundred forty eight lambs showed clinical signs of intoxication and 14 died. The clinical signs included mydriasis, nystagmus, and negative pupillary reflex, bilateral blindness, bump into objects, and lateral movement of the head. No macroscopic lesions were observed. The histological lesions of the retina were cytoplasmic vacuolization in ganglion cells and in cells of the inner and outer nuclear layers with different degrees of atrophy. Vacuolization and axonal degeneration were observed in the optic nerve, with multifocal areas of fibrosis and infiltration by lymphocytes and Gitter cells. To reproduce the intoxication, four sheep were given two, four and 10 times the therapeutic dose of Closantel (0.1g/kg of BW). Only the animals receiving 10 times the recommended dose showed clinical signs. The histological examination of the lesions in experimental sheep showed similar results to those described in the accidental outbreak, except for the absence of optic nerve fibrosis and inflammation, characterizing an acute phase. Axonal myelin sheaths loss, fibroblasts and collagen fibers were observed in the ultrastructural study of the optic nerve of accidental intoxicated animals. The optic nerve of experimentally intoxicated animals had vacuoles that separated the myelin sheaths of axons. To prevent outbreaks it is suggested to weigh the animals before Closantel administration to avoid errors in dose calculation.

The neural processes generating visual phenomenal consciousness: ERP and neuronavigated brain stimulation studies

One of the greatest conundrums to the contemporary science is the relation between consciousness and brain activity, and one of the specifi c questions is how neural activity can generate vivid subjective experiences. Studies focusing on visual consciousness have become essential in solving the empirical questions of consciousness. Th e main aim of this thesis is to clarify the relation between visual consciousness and the neural and electrophysiological processes of the brain. By applying electroencephalography and functional magnetic resonance image-guided transcranial magnetic stimulation (TMS), we investigated the links between conscious perception and attention, the temporal evolution of visual consciousness during stimulus processing, the causal roles of primary visual cortex (V1), visual area 2 (V2) and lateral occipital cortex (LO) in the generation of visual consciousness and also the methodological issues concerning the accuracy of targeting TMS to V1. Th e results showed that the fi rst eff ects of visual consciousness on electrophysiological responses (about 140 ms aft er the stimulus-onset) appeared earlier than the eff ects of selective attention, and also in the unattended condition, suggesting that visual consciousness and selective attention are two independent phenomena which have distinct underlying neural mechanisms. In addition, while it is well known that V1 is necessary for visual awareness, the results of the present thesis suggest that also the abutting visual area V2 is a prerequisite for conscious perception. In our studies, the activation in V2 was necessary for the conscious perception of change in contrast for a shorter period of time than in the case of more detailed conscious perception. We also found that TMS in LO suppressed the conscious perception of object shape when TMS was delivered in two distinct time windows, the latter corresponding with the timing of the ERPs related to the conscious perception of coherent object shape. Th e result supports the view that LO is crucial in conscious perception of object coherency and is likely to be directly involved in the generation of visual consciousness. Furthermore, we found that visual sensations, or phosphenes, elicited by the TMS of V1 were brighter than identically induced phosphenes arising from V2. Th ese fi ndings demonstrate that V1 contributes more to the generation of the sensation of brightness than does V2. Th e results also suggest that top-down activation from V2 to V1 is probably associated with phosphene generation. The results of the methodological study imply that when a commonly used landmark (2 cm above the inion) is used in targeting TMS to V1, the TMS-induced electric fi eld is likely to be highest in dorsal V2. When V1 was targeted according to the individual retinotopic data, the electric fi eld was highest in V1 only in half of the participants. Th is result suggests that if the objective is to study the role of V1 with TMS methodology, at least functional maps of V1 and V2 should be applied with computational model of the TMS-induced electric fi eld in V1 and V2. Finally, the results of this thesis imply that diff erent features of attention contribute diff erently to visual consciousness, and thus, the theoretical model which is built up of the relationship between visual consciousness and attention should acknowledge these diff erences. Future studies should also explore the possibility that visual consciousness consists of several processing stages, each of which have their distinct underlying neural mechanisms.

Maternal aggression in Wistar rats: effect of 5-HT2A/2C receptor agonist and antagonist microinjected into the dorsal periaqueductal gray matter and medial septum

The objective of the present study was to assess the role of the 5-HT2A/2C receptor at two specific brain sites, i.e., the dorsal periaqueductal gray matter (DPAG) and the medial septal (MS) area, in maternal aggressive behavior after the microinjection of either a 5-HT2A/2C receptor agonist or antagonist. Female Wistar rats were microinjected on the 7th postpartum day with the selective agonist alpha-methyl-5-hydroxytryptamine maleate (5-HT2A/2C) or the antagonist 5-HT2A/2C, ketanserin. The agonist was injected into the DPAG at 0.2 (N = 9), 0.5 (N = 10), and 1.0 µg/0.2 µl (N = 9), and the antagonist was injected at 1.0 µg/0.2 µl (N = 9). The agonist was injected into the medial septal area (MS) at 0.2 (N = 9), 0.5 (N = 7), and 1.0 µg/0.2 µl (N = 6) and the antagonist was injected at 1.0 µg/0.2 µl (N = 5). For the control, saline was injected into the DPAG (N = 7) and the MS (N = 12). Both areas are related to aggressive behavior and contain a high density of 5-HT receptors. Non-aggressive behaviors such as horizontal locomotion (walking) and social investigation and aggressive behaviors such as lateral threat (aggressive posture), attacks (frontal and lateral), and biting the intruder were analyzed when a male intruder was placed into the female resident's cage. For each brain area studied, the frequency of the behaviors was compared among the various treatments by analysis of variance. The results showed a decrease in maternal aggressive behavior (number of bites directed at the intruder) after microinjection of the agonist at 0.2 and 1.0 µg/0.2 µl (1.6 ± 0.7 and 0.9 ± 0.3) into the DPAG compared to the saline group (5.5 ± 1.1). There was no dose-response relationship with the agonist. The present findings suggest that the 5-HT2A/2C receptor agonist has an inhibitory effect on maternal aggressive behavior when microinjected into the DPAG and no effect when microinjected into the MS. Ketanserin (1.0 µg/0.2 µl) decreased locomotion when microinjected into the DPAG and MS, but did not affect aggressive behavior. We interpret these findings as evidence for a specific role of 5-HT2A/2C receptors in the DPAG in the inhibition of female aggressive behavior, dissociated from those on motor activity.

Interactions between the cholinergic and dopaminergic systems during the production of ultrasonic vocalizations in rats

Rats produce ultrasonic vocalizations that can be categorized into two types of ultrasonic calls based on their sonographic structure. One group contains 22-kHz ultrasonic vocalization (USVs), characterized by relatively constant (flat) frequency with peak frequency ranging from 19 to 28-kHz, and a call duration ranging between 100 – 3000 ms. These vocalization can be induced by cholinomimetic agents injected into the ascending mesolimbic cholinergic system that terminates in the anterior hypothalamic-preoptic area (AH-MPO) and lateral septum (LS). The other group of USVs contains 50-kHz USVs, characterized by high peak frequency, ranging from 39 to 90-kHz, short duration ranging from 10-90 ms, and varying frequency and complex sonographic morphology. These vocalizations can be induced by dopaminergic agents injected into the nucleus accumbens, the target area for the mesolimbic dopaminergic system. 22-kHz USVs are emitted in situations that are highly aversive, such as proximity of a predator or anticipation of a foot shock, while 50 kHz USVs are emitted in rewarding and appetitive situations, such as juvenile play behaviour or anticipation of rewarding electrical brain stimulation. The activities of these two mesolimbic systems were postulated to be antagonistic to each other. The current thesis is focused on the interaction of these systems indexed by emission of relevant USVs. It was hypothesized that emission of 22 kHz USVs will be antagonized by prior activation of the dopaminergic system while emission of 50 kHz will be antagonized by prior activation of the cholinergic system. It was found that injection of apomorphine into the shell of the nucleus accumbens significantly decreased the number of carbachol-induced 22 kHz USVs from both AH-MPO and LS. Injection of carbachol into the LS significantly decreased the number of apomorphine-induced 50 kHz USVs from the shell of the nucleus accumbens. The results of the study supported the main hypotheses that the mesolimbic dopaminergic and cholinergic systems function in antagonism to each other.

Interactions between the cholinergic and dopaminergic system during the production of ultrasonic vocalizations in rats

Rats produce ultrasonic vocalizations that can be categorized into two types of ultrasonic calls based on their sonographic structure. One group contains 22-kHz ultrasonic vocalization (USVs), characterized by relatively constant (flat) frequency with peak frequency ranging from 19 to 28-kHz, and a call duration ranging between 100 – 3000 ms. These vocalization can be induced by cholinomimetic agents injected into the ascending mesolimbic cholinergic system that terminates in the anterior hypothalamic-preoptic area (AH-MPO) and lateral septum (LS). The other group of USVs contains 50-kHz USVs, characterized by high peak frequency, ranging from 39 to 90-kHz, short duration ranging from 10-90 ms, and varying frequency and complex sonographic morphology. These vocalizations can be induced by dopaminergic agents injected into the nucleus accumbens, the target area for the mesolimbic dopaminergic system. 22-kHz USVs are emitted in situations that are highly aversive, such as proximity of a predator or anticipation of a foot shock, while 50 kHz USVs are emitted in rewarding and appetitive situations, such as juvenile play behaviour or anticipation of rewarding electrical brain stimulation. The activities of these two mesolimbic systems were postulated to be antagonistic to each other. The current thesis is focused on the interaction of these systems indexed by emission of relevant USVs. It was hypothesized that emission of 22 kHz USVs will be antagonized by prior activation of the dopaminergic system while emission of 50 kHz will be antagonized by prior activation of the cholinergic system. It was found that injection of apomorphine into the shell of the nucleus accumbens significantly decreased the number of carbachol-induced 22 kHz USVs from both AH-MPO and LS. Injection of carbachol into the LS significantly decreased the number of apomorphine-induced 50 kHz USVs from the shell of the nucleus accumbens. The results of the study supported the main hypotheses that the mesolimbic dopaminergic and cholinergic systems function in antagonism to each other.

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.

Modelling the growth and movement of cyanobacteria in river systems

Bloom-forming and toxin-producing cyanobacteria remain a persistent nuisance across the world. Modelling of cyanobacteria in freshwaters is an important tool for understanding their population dynamics and predicting the location and timing of the bloom events in lakes and rivers. In this article, a new deterministic model is introduced which simulates the growth and movement of cyanobacterial blooms in river systems. The model focuses on the mathematical description of the bloom formation, vertical migration and lateral transport of colonies within river environments by taking into account the four major factors that affect the cyanobacterial bloom formation in freshwaters: light, nutrients, temperature and river flow. The model consists of two sub-models: a vertical migration model with respect to growth of cyanobacteria in relation to light, nutrients and temperature; and a hydraulic model to simulate the horizontal movement of the bloom. This article presents the model algorithms and highlights some important model results. The effects of nutrient limitation, varying illumination and river flow characteristics on cyanobacterial movement are simulated. The results indicate that under high light intensities and in nutrient-rich waters colonies sink further as a result of carbohydrate accumulation in the cells. In turbulent environments, vertical migration is retarded by vertical velocity component generated by turbulent shear stress. (c) 2006 Elsevier B.V. All rights reserved.