Refine
Year of publication
- 2010 (5) (remove)
Document Type
- Doctoral Thesis (5) (remove)
Language
- English (5)
Has Fulltext
- yes (5)
Is part of the Bibliography
- no (5)
Keywords
- Conformational transitions (1)
- Delaunay-Triangulierung (1)
- Entwicklungspsychologie (1)
- Gedächtnis (1)
- Gedächtnisbildung (1)
- Großhirnrinde (1)
- Konformationsübergänge (1)
- Langzeitgedächtnis (1)
- NREM-Schlaf (1)
- Neural net (1)
Institute
- Frankfurt Institute for Advanced Studies (FIAS) (5) (remove)
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.
At present, there is a huge lag between the artificial and the biological information processing systems in terms of their capability to learn. This lag could be certainly reduced by gaining more insight into the higher functions of the brain like learning and memory. For instance, primate visual cortex is thought to provide the long-term memory for the visual objects acquired by experience. The visual cortex handles effortlessly arbitrary complex objects by decomposing them rapidly into constituent components of much lower complexity along hierarchically organized visual pathways. How this processing architecture self-organizes into a memory domain that employs such compositional object representation by learning from experience remains to a large extent a riddle. The study presented here approaches this question by proposing a functional model of a self-organizing hierarchical memory network. The model is based on hypothetical neuronal mechanisms involved in cortical processing and adaptation. The network architecture comprises two consecutive layers of distributed, recurrently interconnected modules. Each module is identified with a localized cortical cluster of fine-scale excitatory subnetworks. A single module performs competitive unsupervised learning on the incoming afferent signals to form a suitable representation of the locally accessible input space. The network employs an operating scheme where ongoing processing is made of discrete successive fragments termed decision cycles, presumably identifiable with the fast gamma rhythms observed in the cortex. The cycles are synchronized across the distributed modules that produce highly sparse activity within each cycle by instantiating a local winner-take-all-like operation. Equipped with adaptive mechanisms of bidirectional synaptic plasticity and homeostatic activity regulation, the network is exposed to natural face images of different persons. The images are presented incrementally one per cycle to the lower network layer as a set of Gabor filter responses extracted from local facial landmarks. The images are presented without any person identity labels. In the course of unsupervised learning, the network creates simultaneously vocabularies of reusable local face appearance elements, captures relations between the elements by linking associatively those parts that encode the same face identity, develops the higher-order identity symbols for the memorized compositions and projects this information back onto the vocabularies in generative manner. This learning corresponds to the simultaneous formation of bottom-up, lateral and top-down synaptic connectivity within and between the network layers. In the mature connectivity state, the network holds thus full compositional description of the experienced faces in form of sparse memory traces that reside in the feed-forward and recurrent connectivity. Due to the generative nature of the established representation, the network is able to recreate the full compositional description of a memorized face in terms of all its constituent parts given only its higher-order identity symbol or a subset of its parts. In the test phase, the network successfully proves its ability to recognize identity and gender of the persons from alternative face views not shown before. An intriguing feature of the emerging memory network is its ability to self-generate activity spontaneously in absence of the external stimuli. In this sleep-like off-line mode, the network shows a self-sustaining replay of the memory content formed during the previous learning. Remarkably, the recognition performance is tremendously boosted after this off-line memory reprocessing. The performance boost is articulated stronger on those face views that deviate more from the original view shown during the learning. This indicates that the off-line memory reprocessing during the sleep-like state specifically improves the generalization capability of the memory network. The positive effect turns out to be surprisingly independent of synapse-specific plasticity, relying completely on the synapse-unspecific, homeostatic activity regulation across the memory network. The developed network demonstrates thus functionality not shown by any previous neuronal modeling approach. It forms and maintains a memory domain for compositional, generative object representation in unsupervised manner through experience with natural visual images, using both on- ("wake") and off-line ("sleep") learning regimes. This functionality offers a promising departure point for further studies, aiming for deeper insight into the learning mechanisms employed by the brain and their consequent implementation in the artificial adaptive systems for solving complex tasks not tractable so far.
The goal of this project is to develop a framework for a cell that takes in consideration its internal structure, using an agent-based approach. In this framework, a cell was simulated as many sub-particles interacting to each other. This sub-particles can, in principle, represent any internal structure from the cell (organelles, etc). In the model discussed here, two types of sub-particles were used: membrane sub-particles and cytosolic elements. A kinetic and dynamic Delaunay triangulation was used in order to define the neighborhood relations between the sub-particles. However, it was soon noted that the relations defined by the Delaunay triangulation were not suitable to define the interactions between membrane sub-particles. The cell membrane is a lipid bilayer, and does not present any long range interactions between their sub-particles. This means that the membrane particles should not be able to interact in a long range. Instead, their interactions should be confined to the two-dimensional surface supposedly formed by the membrane. A method to select, from the original three-dimensional triangulations, connections restricted to the two-dimensional surface formed by the cell membrane was then developed. The algorithm uses as starting point the three-dimensional Delaunay triangulation involving both internal and membrane sub-particles. From this triangulation, only the subset of connections between membrane sub-particles was considered. Since the cell is full of internal particles, the collection of the membrane particles' connections will resemble the surface to be obtained, even though it will still have many connections that do not belong to the restricted triangulation on the surface. This "thick surface" was called a quasi-surface. The following step was to refine the quasi-surface, cutting out some of the connections so that the ones left made a proper surface triangulation with the membrane points. For that, the quasi-surface was separated in clusters. Clusters are defined as areas on the quasi-surface that are not yet properly triangulated on a two-dimensional surface. Each of the clusters was then re-triangulated independently, using re-triangulation methods also developed during this work. The interactions between cytosolic elements was given by a Lennard-Jones potential, as well as the interactions between cytosolic elements and membrane particles. Between only membrane particles, the interactions were given by an elastic interaction. For each particle, the equation of motion was written. The algorithm chosen to solve the equations of motion was the Verlet algorithm. Since the cytosol can be approximated as a gel, it is reasonable to suppose that the sub-cellular particles are moving in an overdamped environment. Therefore, an overdamped approximation was used for all interactions. Additionally, an adaptive algorithm was used in order to define the size of the time step used in each interaction. After the method to re-triangulate the membrane points was implemented, the time needed to re-triangulate a single cluster was studied, followed by an analysis on how the time needed to re-triangulate each point in a cluster varied with the cluster size. The frequency of appearance for each cluster size was also compared, as this information is necessary to guarantee that the total time needed by to re-triangulate a cell is convergent. At last, the total time spent re-triangulating a surface was plotted, as well as a scaling for the total re-triangulation time with the variation. Even though there is still a lot to be done, the work presented here is an important step on the way to the main goal of this project: to create an agent-based framework that not only allows the simulation of any sub-cellular structure of interest but also provides meaningful interaction relations to particles belonging to the cell membrane.
In the present work, the problem of protein folding is addressed from the point of view of equilibrium thermodynamics. The conformation of a globular protein in solution at common temperatures is quite complicated without any geometrical symmetry, but it is an ordered state in the sense of its biological activity. This complicated conformation of a single protein molecule is destroyed upon increasing the temperature or by the addition of appropriate chemical agents, as is revealed by the loss of its activity and change of the physical properties, and so on. Once the complicated native structures having biological activity are lost, it would be natural to suppose that the native structure could hardly be restored. Nevertheless, pioneers, such as Anson and Mirsky, recognized as early as in 1925 that this was not always the case. If one defines the folded and unfolded states of a protein as two distinct phases of a system, then under the variation of temperature the system is transformed from one phase state into another and vice versa. The process of protein folding is accompanied by the release or absorption of a certain amount of energy, corresponding to the first-oder-type phase transitions in the bulk. Knowing the partition function of the system one can evaluate its energy and heat capacity under different temperatures. This task was performed in this work. The results of the developed statistical mechanics model were compared with the results of molecular dynamic simulations of alanine poylpeptides. In particular, the dependencies on temperature of the total energy of the system and heat capacity were compared for alanine polypeptides consisting of 21, 30, 40, 50 and 100 amino acids. The good correspondence of the results of the theoretical model with the results of molecular dynamics simulations allowed to validate the assumptions made about the system and to establish the accuracy range of the theory. In order to perform the comparison of the results of theoretical model and the molecular dynamics simulations it is necessary to perform the efficient analysis of the results of molecular dynamics simulations. This task was also addressed in the present work. In particular, different ways to obtain dependence of the heat capacity on temperature from molecular dynamics simulations are discussed and the most efficient one is proposed. The present thesis reports the result of molecular dynamic simulations for not only alanine polypeptides by also for valine and leucine polypeptides. In valine and leucine polypeptides, it is also possible to observe the helix↔random coil transitions with the increase of temperature. The current thesis presents a work that starts with the investigation of the fundamental degrees of freedom in polypeptides that are responsible for the conformational transitions. Then this knowledge is applied for the statistical mechanics description of helix↔coil transitions in polypeptides. Finally, the theoretical formalism is generalized for the case of proteins in water environment and the comparison of the results of the statistical mechanics model with the experimental measurements of the heat capacity on temperature dependencies for two globular proteins is performed. The presented formalism is based on fundamental physical properties of the system and provides the possibility to describe the folding↔unfolding transitions quantitatively. The combination of these two facts is the major novelty of the presented approach in comparison to the existing ones. The “transparent” physical nature of the formalism provides a possibility to further apply it to a large variety of systems and processes. For instance, it can be used for investigation of the influence of the mutations in the proteins on their stability. This task is of primary importance for design of novel proteins and drug delivering molecules in medicine. It can provide further insights into the problem of protein aggregation and formation of amyloids. The problem of protein aggregation is closely associated with various illnesses such as Alzheimer and mad cow disease. With certain modifications, the presented theoretical method can be applied to the description of the protein crystallization process, which is important for the determination of the structure of proteins with X-Rays. There many other possible applications of the ideas described in the thesis. For instance, the similar formalism can be developed for the description of melting and unzipping of DNA, growth of nanotubes, formation of fullerenes, etc.
This thesis is dedicated to the study of fluctuation and correlation observables of hadronic equilibrium systems. The statistical hadronization model of high energy physics, in its ideal, i.e. non-interacting, gas approximation will be investigated in different ensemble formulations. The hypothesis of thermal and chemical equilibrium in high energy interaction will be tested against qualitative and quantitative predictions.