Refine
Year of publication
Document Type
- Doctoral Thesis (23)
Has Fulltext
- yes (23)
Is part of the Bibliography
- no (23)
Keywords
- Großhirnrinde (2)
- Schlaf (2)
- Sehrinde (2)
- sparse coding (2)
- Abstandsinformation (1)
- Auditory cortex (1)
- Bilderkennung (1)
- EEG (1)
- Entwicklungspsychologie (1)
- Gedächtnis (1)
Institute
- Physik (12)
- Informatik (7)
- Frankfurt Institute for Advanced Studies (FIAS) (3)
- Informatik und Mathematik (3)
- Medizin (1)
Inhaltsverzeichnis 1. Einleitung …………………………………………………………………...3 1.1 Erklärungsversuche und Forschungsergebnisse der Gegenwart ……8 1.2 Zielrichtung und Abgrenzung der aktuellen Arbeit ………………..21 1.3 Intention und Erläuterung der Versuchsreihen ………………….....25 2. Grundlagen und Methodiken bezüglich des subjektiven visuellen Wahrnehmungsraums …………………………………………………........27 2.1 Die nativistische und die empiristische Anschauung ………………27 2.2 Räumliche Anordnungen der wahrgenommenen Objekte …………31 2.3 Über die visuell vermittelte Richtungs- und Lagebestimmung …....33 2.4 Visuelle Auswertungen der korrespondierenden Netzhautstellen …42 2.5 Visuelle Auswertungen der disparaten Netzhautstellen …………...44 2.6 Die Größenkonstanzleistung ………………………………………47 2.7 Psychophysikalische Grundlagen und Schwellenwerte …………...50 2.8 Physiologische Grundlagen ………………………………………..54 3. Experimentelle Untersuchung ……………………………………………..60 3.1 Versuchsaufbau und Ablauf zur Durchführung der Experimente …60 3.1.1 Zusammensetzungen der Versuchsteilnehmer ……………66 3.1.2 Erläuterungen und Ablauf der 2 Versuchsreihen …………66 3.2 Graphische Darstellungen der Messergebnisse ……………………71 3.2.1 I.Versuchsreihe ……………………………………………71 3.2.2 II.Versuchsreihe …………………………………………...93 3.3 Auswertung und Aufbereitung der Messdaten …………………..102 3.3.1 Auswertungen der I.Versuchsreihe ……………………..102 3.3.2 Auswertungen der II.Versuchsreihe …………………….120 3.3.3 Fehlerbetrachtungen der Versuchsreihen I und II ………122 3.4 Diskussion der Messdaten ……………………………………….124 4. Zusammenfassung und Ausblick ………………………………………...135 Begriffsverzeichnis mit kurzer Erklärung.…………………………………...137 Literaturverzeichnis …………………………………………………….........141 Bildquellenverzeichnis ………………………………………………….......143 Als Fazit kann man folgendes zusammenfassend sagen: Die aufgestellte Arbeitshypothese wurde durch die beiden Versuchsreihen verifiziert, denn die Ergebnisse ergaben folgendes: - In den Messreihen der Versuchsreihe I ist jeweils ein Anstieg der eingestellten Größe, je mehr Abstandsinformationen zugelassen wurden, zu verzeichnen. Das bedeutet, der Anstieg wurde umso größer, desto größer die AID wurde. Auch waren in allen Messreihen die monokularen Größeneinstellungen, bei sonst konstanter AID, gegenüber der binokularen Größeneinstellung geringer. Bei Verringerung der Einstellentfernung wurden die Abweichungen zwischen den subjektiven und den objektiven Größen ebenfalls größer. Das heißt also, die subjektive visuelle Wahrnehmungsgröße ist von der AID wie folgt abhängig: Das visuelle System bewertet subjektiv die Wahrnehmungsgröße bei maximaler AID nach oben und relativ dazu, bei minimaler AID nach unten. - Dass die aufgestellten Parameter die AID bedingen, konnte durch die 1. Messreihe gezeigt werden, da der jeweilige Anstieg der eingestellten Größe, nur durch die Variation eines Parameters erfolgte. Die Querdisparation konnte aber hier nicht als Parameter der die AID bedingt isoliert untersucht werden. Bei den meisten Probanden ergaben sich sehr schnell Doppelbilder und erzeugten bei ihnen ein Unbehagen. Aber dennoch floss dieser Parameter als einflussnehmende Größe in den Konvergenzgrad mit ein. Das Netzhautbild konnte nur kombiniert mit dem psychologischen Gefühl der Nähe isoliert betrachtet werden. Damit die Voraussetzungen in beiden Versuchen gleich waren, wurde in der Versuchsreihe II unter gleichen Versuchsbedingungen wie in der Versuchsreihe I gemessen. Auch hier wurden die Abstandsinformationen von minimal bis maximal sukzessive zugelassen. Durch die Messdaten der Versuchsreihe II konnte eindeutig gezeigt werden, dass die Abstandsunterschiedsschwelle umso geringer ausfällt, desto mehr Abstandskriterien hinzukamen, also die AID erhöht wurde. Analog kehren sich die Verhältnisse um, wenn die AID erniedrigt wird. Durch diesen kausalen Zusammenhang zwischen der Abstandsunterschiedsschwelle des visuellen Systems und der Güte der AID bestätigt sich zusätzlich die Annahme, dass die eingeführten Parameter des Abstandes tatsächlich als solche zu betrachten sind und die AID konstituieren. Denn wären sie keine Konstituenten der AID, so müssten die Unterschiedsschwellen der Versuchsreihe II in etwa gleich sein. Da aber die Änderung der Randbedingungen sich auf die verwertbaren Abstandsinformationen bezogen und somit die AID jeweils geändert wurde, ist die aufgestellte Annahme über die Parameter, welche die AID bedingen, berechtigt. - Dass im orthostereoskopischen Bereich die subjektiven Größeneinstellungen gegenüber der Zentralprojektion am weitesten auseinander liegen, bestätigte sich durch alle Messreihen der Versuchsreihe I. In diesem Bereich existiert die maximale Unabhängigkeit der visuellen Wahrnehmungsgröße vom Gesichtswinkel. In diesem Bereich liegt eine sehr hohe Güte in der Größenkonstanzleistung des visuellen Systems vor. Dass die Größenkonstanz qualitativ dem aufgestellten Formalismus aus Annahme 2 genügt und die aufgestellte qualitative Relation sie beschreibt, konnte nicht gezeigt werden. Das begründet sich durch das Zustandekommen der Größenkonstanz. Sie resultiert bekanntlich aus einer Entfernungsänderung. Je nach dem, ob sich ein Objekt dem Beobachter nähert oder entfernt, setzt diese Bildgrößenkompensation ein. Von daher unterliegt sie einem dynamischen Prozess und kann dadurch mit Relation (2´) nicht beschreiben werden. - Mit der Relation 2´ kann man qualitativ die Unbestimmtheit in der visuellen Wahrnehmungsentfernung beschreiben und qualitativ erklären. Der Aspekt der Abstandsunterschiedsschwelle ist etwas verwirrend. Auf der einen Seite handelt es sich um eine Vermögensleistung des visuellen Systems, welches abhängig ist von den vorliegenden Abstandsinformationen, die ihrerseits die AID bedingen. Auf der anderen Seite bedingt die Abstandsunterschiedsschwelle die AID durch ihre Güte und Qualität, beeinflusst also umgekehrt auch die AID. In der Versuchsreihe 2 wurde auf die Vermögensleistung des visuellen Systems und deren Abhängigkeit von den Parametern eingegangen, die auch die AID bedingen. Dies diente dazu, zusätzlich zu zeigen, dass es sich bei diesen Parametern um Parameter handelt, welche die AID bedingen. Die Argumentationskette lautete wie folgt: Die Abstandsunterschiedsschwelle beeinflusst die AID. Die betrachteten Parameter beeinflussten die Abstandsunterschiedsschwelle, dass experimentell verifiziert wurde. Daraus folgte dann, dass eben diese Parameter auch die AID bedingen. Diese Argumentation diente nur als zusätzliches Hilfsmittel. Bei Punkt 4 sollte die Abstandsunterschiedsschwelle und ihr Einfluss auf die Unbestimmtheit hin betrachtet werden. Dies hat aber nur sekundäre Relevanz, da hier die Anwendung der Relation 2´ im Vordergrund stand. - Ob die Fitting-Funktion, welche die Messdaten der Versuchsreihe I approximierte, sich als Algorithmus für die Darstellung einer Bewegungssimulation eignet, kann noch nicht gesagt werden. Es müssen noch Untersuchungen umgesetzt werden, welche die Diagonalbewegung beschreiben. In der stirnfrontalen Vor- und Zurückbewegung ist der simulierte Bewegungsablauf mit der Fitting-Funktion gegenüber der linearen Darstellung realistischer. Dies ist in der ersten 100cm Raumtiefe besonders merklich, da die Fitting-Funktion die Größenkonstanzleistung des visuellen Systems berücksichtigt. Die auf dem konventionellen Computerspielmarkt eingesetzten Algorithmen für die Darstellung von Vor- und Zurückbewegungen sind dagegen nahezu linear, welches dem Beobachter einen etwas unnatürlichen Seheindruck vermittelt. Die Fitting-Funktion könnte auch für die Simulation von Zeichentrickfilmen verwendet werden. Auch dort wird die Größenkonstanzleistung des visuellen Systems nicht berücksichtigt. Aber gerade diese Konstanzleistung gestaltet die Größenvariation der wahrgenommenen Objekte bei Entfernungsänderungen. Dies ist besonders im orthostereoskopischen Bereich merklich.
Already today modern driver assistance systems contribute more and more to make individual mobility in road traffic safer and more comfortable. For this purpose, modern vehicles are equipped with a multitude of sensors and actuators which perceive, interpret and react to the environment of the vehicle. In order to reach the next set of goals along this path, for example to be able to assist the driver in increasingly complex situations or to reach a higher degree of autonomy of driver assistance systems, a detailed understanding of the vehicle environment and especially of other moving traffic participants is necessary.
It is known that motion information plays a key role for human object recognition [Spelke, 1990]. However, full 3D motion information is mostly not taken into account for Stereo Vision-based object segmentation in literature. In this thesis, novel approaches for motion-based object segmentation of stereo image sequences are proposed from which a generic environmental model is derived that contributes to a more precise analysis and understanding of the respective traffic scene. The aim of the environmental model is to yield a minimal scene description in terms of a few moving objects and stationary background such as houses, crash barriers or parking vehicles. A minimal scene description aggregates as much information as possible and it is characterized by its stability, precision and efficiency.
Instead of dense stereo and optical flow information, the proposed object segmentation builds on the so-called Stixel World, an efficient superpixel-like representation of space-time stereo data. As it turns out this step substantially increases stability of the segmentation and it reduces the computational time by several orders of magnitude, thus enabling real-time automotive use in the first place. Besides the efficient, real-time capable optimization, the object segmentation has to be able to cope with significant noise which is due to the measurement principle of the used stereo camera system. For that reason, in order to obtain an optimal solution under the given extreme conditions, the segmentation task is formulated as a Bayesian optimization problem which allows to incorporate regularizing prior knowledge and redundancies into the object segmentation.
Object segmentation as it is discussed here means unsupervised segmentation since typically the number of objects in the scene and their individual object parameters are not known in advance. This information has to be estimated from the input data as well.
For inference, two approaches with their individual pros and cons are proposed, evaluated and compared. The first approach is based on dynamic programming. The key advantage of this approach is the possibility to take into account non-local priors such as shape or object size information which is impossible or which is prohibitively expensive with more local, conventional graph optimization approaches such as graphcut or belief propagation.
In the first instance, the Dynamic Programming approach is limited to one-dimensional data structures, in this case to the first Stixel row. A possible extension to capture multiple Stixel rows is discussed at the end of this thesis.
Further novel contributions include a special outlier concept to handle gross stereo errors associated with so-called stereo tear-off edges. Additionally, object-object interactions are taken into account by explicitly modeling object occlusions. These extensions prove to be dramatic improvements in practice.
This first approach is compared with a second approach that is based on an alternating optimization of the Stixel segmentation and of the relevant object parameters in an expectation maximization (EM) sense. The labeling step is performed by means of the _−expansion graphcut algorithm, the parameter estimation step is done via one-dimensional sampling and multidimensional gradient descent. By using the Stixel World and due to an efficient implementation, one step of the optimization only takes about one millisecond on a standard single CPU core. To the knowledge of the author, at the time of development there was no faster global optimization in a demonstrator car.
For both approaches, various testing scenarios have been carefully selected and allow to examine the proposed methods thoroughly under different real-world conditions with limited groundtruth at hand. As an additional innovative application, the first approach was successfully implemented in a demonstrator car that drove the so-called Bertha Benz Memorial Route from Mannheim to Pforzheim autonomously in real traffic.
At the end of this thesis, the limits of the proposed systems are discussed and a prospect on possible future work is given.
This thesis investigates the development of early cognition in infancy using neural network models. Fundamental events in visual perception such as caused motion, occlusion, object permanence, tracking of moving objects behind occluders, object unity perception and sequence learning are modeled in a unifying computational framework while staying close to experimental data in developmental psychology of infancy. In the first project, the development of causality and occlusion perception in infancy is modeled using a simple, three-layered, recurrent network trained with error backpropagation to predict future inputs (Elman network). The model unifies two infant studies on causality and occlusion perception. Subsequently, in the second project, the established framework is extended to a larger prediction network that models the development of object unity, object permanence and occlusion perception in infancy. It is shown that these different phenomena can be unified into a single theoretical framework thereby explaining experimental data from 14 infant studies. The framework shows that these developmental phenomena can be explained by accurately representing and predicting statistical regularities in the visual environment. The models assume (1) different neuronal populations processing different motion directions of visual stimuli in the visual cortex of the newborn infant which are supported by neuroscientific evidence and (2) available learning algorithms that are guided by the goal of predicting future events. Specifically, the models demonstrate that no innate force notions, motion analysis modules, common motion detectors, specific perceptual rules or abilities to "reason" about entities which have been widely postulated in the developmental literature are necessary for the explanation of the discussed phenomena. Since the prediction of future events turned out to be fruitful for theoretical explanation of various developmental phenomena and a guideline for learning in infancy, the third model addresses the development of visual expectations themselves. A self-organising, fully recurrent neural network model that forms internal representations of input sequences and maps them onto eye movements is proposed. The reinforcement learning architecture (RLA) of the model learns to perform anticipatory eye movements as observed in a range of infant studies. The model suggests that the goal of maximizing the looking time at interesting stimuli guides infants' looking behavior thereby explaining the occurrence and development of anticipatory eye movements and reaction times. In contrast to classical neural network modelling approaches in the developmental literature, the model uses local learning rules and contains several biologically plausible elements like excitatory and inhibitory spiking neurons, spike-timing dependent plasticity (STDP), intrinsic plasticity (IP) and synaptic scaling. It is also novel from the technical point of view as it uses a dynamic recurrent reservoir shaped by various plasticity mechanisms and combines it with reinforcement learning. The model accounts for twelve experimental studies and predicts among others anticipatory behavior for arbitrary sequences and facilitated reacquisition of already learned sequences. All models emphasize the development of the perception of the discussed phenomena thereby addressing the questions of how and why this developmental change takes place - questions that are difficult to be assessed experimentally. Despite the diversity of the discussed phenomena all three projects rely on the same principle: the prediction of future events. This principle suggests that cognitive development in infancy may largely be guided by building internal models and representations of the visual environment and using those models to predict its future development.
This dissertation connects two independent fields of theoretical neuroscience: on the one hand, the self-organization of topographic connectivity patterns, and on the other hand, invariant object recognition, that is the recognition of objects independently of their various possible retinal representations (for example due to translations or scalings). The topographic representation is used in the presented approach, as a coordinate system, which then allows for the implementation of invariance transformations. Hence this study shows, that it is possible that the brain self-organizes before birth, so that it is able to invariantly recognize objects immediately after birth. Besides the core hypothesis that links prenatal work with object recognition, advancements in both fields themselves are also presented. In the beginning of the thesis, a novel analytically solvable probabilistic generative model for topographic maps is introduced. And at the end of the thesis, a model that integrates classical feature-based ideas with the normalization-based approach is presented. This bilinear model makes use of sparseness as well as slowness to implement "optimal" topographic representations. It is therefore a good candidate for hierarchical processing in the brain and for future research.
The brain is a highly dynamic and variable system: when the same stimulus is presented to the same animal on the same day multiple times, the neural responses show high trial-to-trial variability. In addition, even in the absence of sensory stimulation neural recordings spontaneously show seemingly random activity patterns. Evoked and spontaneous neural variability is not restricted to activity but is also found in structure: most synapses do not survive for longer than two weeks and even those that do show high fluctuations in their efficacy.
Both forms of variability are further affected by stochastic components of neural processing such as frequent transmission failure. At present it is unclear how these observations relate to each other and how they arise in cortical circuits.
Here, we will investigate how the self-organizational processes of neural circuits affect the high variability in two different directions: First, we will show that recurrent dynamics of self-organizing neural networks can account for key features of neural variability. This is achieved in the absence of any intrinsic noise sources by the neural network models learning a predictive model of their environment with sampling-like dynamics. Second, we will show that the same self-organizational processes can compensate for intrinsic noise sources. For this, an analytical model and more biologically plausible models are established to explain the alignment of parallel synapses in the presence of synaptic failure.
Both modeling studies predict properties of neural variability, of which two are subsequently tested on a synapse database from a dense electron microscopy reconstruction from mouse somatosensory cortex and on multi-unit recordings from the visual cortex of macaque monkeys during a passive viewing task. While both analyses yield interesting results, the predicted properties were not confirmed, guiding the next iteration of experiments and modeling studies.
Cortical circuits exhibit highly dynamic and complex neural activity. Intriguingly, cortical activity exhibits consistently two key features across observed species and brain areas. First, individual neurons tend to be co-active in spatially localized domains forming orderly arranged, modular layouts with a typical spatial scale. Second, cortical elements are correlated in their activity over large distances reflecting long-range network interactions distributed over several millimeters. Currently, it is unclear how these two fundamental properties emerge in the early developing cortical activity.
Here, I aim to fill this gap by combining analyses of chronic imaging data and network models of developing cortical activity. Neural recordings of spontaneous and visually evoked activity in primary visual cortex of ferrets during their early cortical development were obtained using in vivo 2-photon and widefield epi-fluorescence calcium imaging. Spontaneous activity was used to probe the early state of cortical networks as its spatiotemporal organization is independent of a stimulus-imposed structure, and it is already present early in cortical development prior to reliably evoked responses. To assess the mature functional organization of distributed networks in cortex, the tuning of neural responses to stimulus features, in particular to the orientation of an edge-like stimulus, was assessed. Cortical responses to moving gratings of varying orientations form an orderly arranged layout of orientation domains extending over several millimeters.
To begin with, I showed that spontaneous activity correlations extend over several millimeters, supporting the assumption of using spontaneous activity to assess distributed networks in cortex.
Next, I asked how distributed networks in the mature visual cortex - assessed by spontaneous activity correlations - are related to its fine-scale functional organization. I found that the spatially extended and modular spontaneous correlation patterns accurately predict the fine spatial structure of visually evoked orientation domains several millimeters away. These results suggest a close relation between spontaneous correlations and visually evoked responses on a fine spatial scale and across large spatial distances.
As the principles governing the functional organization and development of distributed network interactions in the neocortex remain poorly understood, I next asked how long range correlated activity arises early in development. I found that key features of mature spontaneous activity introduced in this work, including long-range spontaneous correlations, were present already early in cortical development prior to the maturation of long-range, horizontal connections, and the predicted mature orientation preference layout. Even after silencing feed-forward input drive by inactivating retina or thalamus, long-range correlated and modular activity robustly emerged in early cortex. These results suggest that local recurrent connections in early cortical circuits can generate structured long-range network correlations that guide the formation of visually-evoked distributed functional networks.
To investigate how these large-scale cortical networks emerge prior to the maturation and elaboration of long-range horizontal connectivity, I examined a statistical network model describing an ensemble of spatially extended spontaneous activity patterns. I found a direct relationship between the dimensionality of this ensemble of activity patterns and the decay of its correlation structure. Specifically, reducing the dimensionality of the ensemble leads to an increase in the spatial range of the correlation structure.
To test whether this mechanism could generate a long-range correlation structure in cortical circuits, I studied a dynamical network model implementing a dimensionality reduction mechanism. Based on previous work demonstrating that network heterogeneity reduces the dimensionality of activity patterns, I showed that by increasing the degree of heterogeneity in the network, the dimensionality of the ensemble of activity patterns decreases and in turn their correlations extend over a greater range. A comparison to experimental data revealed a quantitative match between the network model and the observations in vivo in several of the key features of the early cortex including the spatial scale of correlations. Low dimensionality of spontaneous activity thus might provide an organizational principle explaining the observed long-range correlation structure in the early cortex.
Finally, I asked whether a network with a biologically plausible architecture can generate modular activity. Several classical models showed that modular activity patterns can emerge via an intracortical mechanism involving lateral inhibition. However, this assumption appears to be in conflict with current experimental evidence. Moreover, these network models were not experimentally tested, so far. Here, I showed by using linear stability analysis that spatially localized self-inhibition relaxes the constraints on the connectivity structure in a network model, such that biologically more plausible network motifs with shorter ranging inhibition than excitation can robustly generate modular activity.
Importantly, I also provided several model predictions to make the class of network models experimentally testable in view of recent technological advancements in imaging and manipulation of cortical circuits. A critical prediction of the model is the decrease in spacing of active domains when the total amount of inhibition increases. These results provide a novel mechanism of how cortical circuits with short-range inhibition can form modular activity.
Taken together, this thesis provides evidence that the two described fundamental features of neural activity are already present in the early cortex and shows that activity with those features can be generated in network models with an architecture consistent with the early cortex using basic principles.
In our daily life, we carry out lots of tasks like typing, playing tennis, and playing the piano, without even noticing there is sequence learning involved. No matter how simple or complex they are, these tasks require the sequential planning and execution of a series of movements. As an ability of primary importance in one’s life, and an ability that everyone manages to learn, action-sequence learning has been studied by researchers from different fields: psychologists, neurophysiologists as well as roboticists. In the concept of sequence learning, perceptual learning and motor learning, implicit and explicit learning have been studied and discussed independently.
We are interested in infancy research, because infants, with underdeveloped brain functions and with limited motor ability, have little experience with the world and not yet built internal models as presumption of how to interpret the world. A series of infant experiments in the 1980s provided evidence that infants can rapidly develop anticipatory eye movements for visual events. Even when infants have no control of those spatial-temporal patterns, they can respond actually prior to the onset of the visual event, referred as "Anticipation".
In this work, we applied a gaze-contingent paradigm using real-time eye tracking to put 6- and 8-month-old infants in direct control of their visual surroundings. This paradigm allows the infant to change an image on a screen by looking at a peripheral red disc, which functions as a switch. We found that infants quickly learn to perform eye movements to trigger the appearance of new stimuli and that they anticipate the consequences of their actions in an early stage of the experiment.
Attention-shift from learning one stimulus to the next novel stimulus is important in sequence learning. In the test phase of infant visual habituation with two objects, we propose a new theory of explaining the familiarity-to-novelty shift. In our opinion an infant’s interest in a stimulus is related to its learning progress, the improvement of performance. As a consequence, infants prefer the stimulus which their current learning progress is maximal for, naturally giving rise to a familiarity-to-novelty shift in certain situations. Our network model predicts that the familiarity-to-novelty-shift only emerges for complex stimuli that produce bell-shaped learning curves after brief familiarization, but does not emerge for simple stimuli that produce exponentially decreasing learning curves or for long familiarization time, which is consistent with experimental results. This research suggests the infant's interest in a stimulus may be related to its current learning progress. This can give rise to a dynamic familiarity-to-novelty shift depending on both the infant's learning efficiency and the task complexity.
We know that for both infants and adults, the performance on certain motor-sequence tasks can be improved through practice. However, adults usually have to perform complex tasks in complicated environments; for example, learning multiple tasks is unavoidable in our daily life. In existing research, learning multiple tasks showed puzzling and seemingly contradictory results. On the one hand, a wide variety of proactive and retroactive interference effects have been observed when multiple tasks have to be learned. On the other hand, some studies have reported facilitation and transfer of learning between different tasks.
In order to find out the interaction between multiple-task learning, and to find an optimal training schedule, we use a recurrent neural network to model a series of experiments on movement sequence learning. The network model learns to carry out the correct movement sequences through training and reproduces differences between training schedules such as blocked training vs. random training in psychophysics experiments. The network model also shows striking similarity to human performance, and makes prediction for tasks similarity and different training schedules.
In conclusion, the thesis presents learning sequences of actions in infants and recurrent neural networks. We carried out a gaze-contingent experiment to study infants’ rapid anticipation of their own action outcomes, and we also constructed two recurrent neural network models, with one model explaining infant attention shift in visual habituation, and the other model directing to task similarity and training schedule in motor sequence control in adults.
Plasticity supports the remarkable adaptability and robustness of cortical processing. It allows the brain to learn and remember patterns in the sensory world, to refine motor control, to predict and obtain reward, or to recover function after injury. Behind this great flexibility hide a range of plasticity mechanisms, affecting different aspects of neuronal communication. However, little is known about the precise computational roles of some of these mechanisms. Here, we show that the interaction between spike-timing dependent plasticity (STDP), intrinsic plasticity and synaptic scaling enables neurons to learn efficient representations of their inputs. In the context of reward-dependent learning, the same mechanisms allow a neural network to solve a working memory task. Moreover, although we make no any apriori assumptions on the encoding used for representing inputs, the network activity resembles that of brain regions known to be associated with working memory, suggesting that reward-dependent learning may be a central force in working memory development. Lastly, we investigated some of the clinical implications of synaptic scaling and showed that, paradoxically, there are situations in which the very mechanisms that normally are required to preserve the balance of the system, may act as a destabilizing factor and lead to seizures. Our model offers a novel explanation for the increased incidence of seizures following chronic inflammation.
Das Gehirn ist die wohl komplexeste Struktur auf Erden, die der Mensch erforscht. Es besteht aus einem riesigen Netzwerk von Nervenzellen, welches in der Lage ist eingehende sensorische Informationen zu verarbeiten um daraus eine sinnvolle Repräsentation der Umgebung zu erstellen. Außerdem koordiniert es die Aktionen des Organismus um mit der Umgebung zu interagieren. Das Gehirn hat die bemerkenswerte Fähigkeit sowohl Informationen zu speichern als auch sich ständig an ändernde Bedingungen anzupassen, und zwar über die gesamte Lebensdauer. Dies ist essentiell für Mensch oder Tier um sich zu entwickeln und zu lernen. Die Grundlage für diesen lebenslangen Lernprozess ist die Plastizität des Gehirns, welche das riesige Netzwerk von Neuronen ständig anpasst und neu verbindet. Die Veränderungen an den synaptischen Verbindungen und der intrinsischen Erregbarkeit jedes Neurons finden durch selbstorganisierte Mechanismen statt und optimieren das Verhalten des Organismus als Ganzes. Das Phänomen der neuronalen Plastizität beschäftigt die Neurowissenschaften und anderen Disziplinen bereits über mehrere Jahrzehnte. Dabei beschreibt die intrinsische Plastizität die ständige Anpassung der Erregbarkeit eines Neurons um einen ausbalancierten, homöostatischen Arbeitsbereich zu gewährleisten. Aber besonders die synaptische Plastizität, welche die Änderungen in der Stärke bestehender Verbindungen bezeichnet, wurde unter vielen verschiedenen Bedingungen erforscht und erwies sich mit jeder neuen Studie als immer komplexer. Sie wird durch ein komplexes Zusammenspiel von biophysikalischen Mechanismen induziert und hängt von verschiedenen Faktoren wie der Frequenz der Aktionspotentiale, deren Timing und dem Membranpotential ab und zeigt außerdem eine metaplastische Abhängigkeit von vergangenen Ereignissen. Letztlich beeinflusst die synaptische Plastizität die Signalverarbeitung und Berechnung einzelner Neuronen und der neuronalen Netzwerke.
Der Schwerpunkt dieser Arbeit ist es das Verständnis der biologischen Mechanismen und deren Folgen, die zu den beobachteten Plastizitätsphänomene führen, durch eine stärker vereinheitlichte Theorie voranzutreiben.Dazu stelle ich zwei funktionale Ziele für neuronale Plastizität auf, leite Lernregeln aus diesen ab und analysiere deren Konsequenzen und Vorhersagen.
Kapitel 3 untersucht die Unterscheidbarkeit der Populationsaktivität in Netzwerken als funktionales Ziel für neuronale Plastizität. Die Hypothese ist dabei, dass gerade in rekurrenten aber auch in vorwärtsgekoppelten Netzwerken die Populationsaktivität als Repräsentation der Eingangssignale optimiert werden kann, wenn ähnliche Eingangssignale eine möglichst unterschiedliche Repräsentation haben und dadurch für die nachfolgende Verarbeitung besser unterscheidbar sind. Das funktionale Ziel ist daher diese Unterscheidbarkeit durch Veränderungen an den Verbindungsstärke und der Erregbarkeit der Neuronen mithilfe von lokalen selbst-organisierten Lernregeln zu maximieren. Aus diesem funktionale Ziel lassen sich eine Reihe von Standard-Lernenregeln für künstliche neuronale Netze gemeinsam abzuleiten.
Kapitel 4 wendet einen ähnlichen funktionalen Ansatz auf ein komplexeres, biophysikalisches Neuronenmodell an. Das Ziel ist eine spärliche, stark asymmetrische Verteilung der synaptischen Stärke, wie sie auch bereits mehrfach experimentell gefunden wurde, durch lokale, synaptische Lernregeln zu maximieren. Aus diesem funktionalen Ansatz können alle wichtigen Phänomene der synaptischen Plastizität erklärt werden. Simulationen der Lernregel in einem realistischen Neuronmodell mit voller Morphologie erklären die Daten von timing-, raten- und spannungsabhängigen Plastizitätsprotokollen. Die Lernregel hat auch eine intrinsische Abhängigkeit von der Position der Synapse, welche mit den experimentellen Ergebnissen übereinstimmt. Darüber hinaus kann die Lernregel ohne zusätzliche Annahmen metaplastische Phänomene erklären. Dabei sagt der Ansatz eine neue Form der Metaplastizität voraus, welche die timing-abhängige Plastizität beeinflusst. Die formulierte Lernregel führt zu zwei neuartigen Vereinheitlichungen für synaptische Plastizität: Erstens zeigt sie, dass die verschiedenen Phänomene der synaptischen Plastizität als Folge eines einzigen funktionalen Ziels verstanden werden können. Und zweitens überbrückt der Ansatz die Lücke zwischen der funktionalen und mechanistische Beschreibungsweise. Das vorgeschlagene funktionale Ziel führt zu einer Lernregel mit biophysikalischer Formulierung, welche mit etablierten Theorien der biologischen Mechanismen in Verbindung gebracht werden kann. Außerdem kann das Ziel einer spärlichen Verteilung der synaptischen Stärke als Beitrag zu einer energieeffizienten synaptischen Signalübertragung und optimierten Codierung interpretiert werden.
This thesis will first introduce in more detail the Bayesian theory and its use in integrating multiple information sources. I will briefly talk about models and their relation to the dynamics of an environment, and how to combine multiple alternative models. Following that I will discuss the experimental findings on multisensory integration in humans and animals. I start with psychophysical results on various forms of tasks and setups, that show that the brain uses and combines information from multiple cues. Specifically, the discussion will focus on the finding that humans integrate this information in a way that is close to the theoretical optimal performance. Special emphasis will be put on results about the developmental aspects of cue integration, highlighting experiments that could show that children do not perform similar to the Bayesian predictions. This section also includes a short summary of experiments on how subjects handle multiple alternative environmental dynamics. I will also talk about neurobiological findings of cells receiving input from multiple receptors both in dedicated brain areas but also primary sensory areas. I will proceed with an overview of existing theories and computational models of multisensory integration. This will be followed by a discussion on reinforcement learning (RL). First I will talk about the original theory including the two different main approaches model-free and model-based reinforcement learning. The important variables will be introduced as well as different algorithmic implementations. Secondly, a short review on the mapping of those theories onto brain and behaviour will be given. I mention the most in uential papers that showed correlations between the activity in certain brain regions with RL variables, most prominently between dopaminergic neurons and temporal difference errors. I will try to motivate, why I think that this theory can help to explain the development of near-optimal cue integration in humans. The next main chapter will introduce our model that learns to solve the task of audio-visual orienting. Many of the results in this section have been published in [Weisswange et al. 2009b,Weisswange et al. 2011]. The model agent starts without any knowledge of the environment and acts based on predictions of rewards, which will be adapted according to the reward signaling the quality of the performed action. I will show that after training this model performs similarly to the prediction of a Bayesian observer. The model can also deal with more complex environments in which it has to deal with multiple possible underlying generating models (perform causal inference). In these experiments I use di#erent formulations of Bayesian observers for comparison with our model, and find that it is most similar to the fully optimal observer doing model averaging. Additional experiments using various alterations to the environment show the ability of the model to react to changes in the input statistics without explicitly representing probability distributions. I will close the chapter with a discussion on the benefits and shortcomings of the model. The thesis continues whith a report on an application of the learning algorithm introduced before to two real world cue integration tasks on a robotic head. For these tasks our system outperforms a commonly used approximation to Bayesian inference, reliability weighted averaging. The approximation is handy because of its computational simplicity, because it relies on certain assumptions that are usually controlled for in a laboratory setting, but these are often not true for real world data. This chapter is based on the paper [Karaoguz et al. 2011]. Our second modeling approach tries to address the neuronal substrates of the learning process for cue integration. I again use a reward based training scheme, but this time implemented as a modulation of synaptic plasticity mechanisms in a recurrent network of binary threshold neurons. I start the chapter with an additional introduction section to discuss recurrent networks and especially the various forms of neuronal plasticity that I will use in the model. The performance on a task similar to that of chapter 3 will be presented together with an analysis of the in uence of different plasticity mechanisms on it. Again benefits and shortcomings and the general potential of the method will be discussed. I will close the thesis with a general conclusion and some ideas about possible future work.