To see the other types of publications on this topic, follow the link: Physics Informed Neural Networks.
Dissertations / Theses on the topic 'Physics Informed Neural Networks'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 dissertations / theses for your research on the topic 'Physics Informed Neural Networks.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
1
Cedergren, Linnéa. "Physics-informed Neural Networks for Biopharma Applications." Thesis, Umeå universitet, Institutionen för fysik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:umu:diva-185423.
Full textAbstract:
Physics-Informed Neural Networks (PINNs) are hybrid models that incorporate differential equations into the training of neural networks, with the aim of bringing the best of both worlds. This project used a mathematical model describing a Continuous Stirred-Tank Reactor (CSTR), to test two possible applications of PINNs. The first type of PINN was trained to predict an unknown reaction rate law, based only on the differential equation and a time series of the reactor state. The resulting model was used inside a multi-step solver to simulate the system state over time. The results showed that the PINN could accurately model the behaviour of the missing physics also for new initial conditions. However, the model suffered from extrapolation error when tested on a larger reactor, with a much lower reaction rate. Comparisons between using a numerical derivative or automatic differentiation in the loss equation, indicated that the latter had a higher robustness to noise. Thus, it is likely the best choice for real applications. A second type of PINN was trained to forecast the system state one-step-ahead based on previous states and other known model parameters. An ordinary feed-forward neural network with an equal architecture was used as baseline. The second type of PINN did not outperform the baseline network. Further studies are needed to conclude if or when physics-informed loss should be used in autoregressive applications.
2
Mirzai, Badi. "Physics-Informed Deep Learning for System Identification of Autonomous Underwater Vehicles : A Lagrangian Neural Network Approach." Thesis, KTH, Skolan för elektroteknik och datavetenskap (EECS), 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-301626.
Full textAbstract:
In this thesis, we explore Lagrangian Neural Networks (LNNs) for system identification of Autonomous Underwater Vehicles (AUVs) with 6 degrees of freedom. One of the main challenges of AUVs is that they have limited wireless communication and navigation under water. AUVs operate under strict and uncertain conditions, where they need to be able to navigate and perform tasks in unknown ocean environments with limited and noisy sensor data. A crucial requirement for localization and adaptive control of AUVs is having an accurate and reliable model of the system’s nonlinear dynamics while taking into account the dynamic environment of the ocean. Most of these dynamics models do not incorporate data. The collection of data for AUVs is difficult, but necessary in order to have more flexibility in the model’s parameters due to the dynamic environment of the ocean. Yet, traditional system identification methods are still dominant today, despite the recent breakthroughs in Deep Learning. Therefore, in this thesis, we aim for a data-driven approach that embeds laws from physics in order to learn the state-space model of an AUV. More precisely, exploring the LNN framework for higher-dimensional systems. Furthermore, we also extend the LNN to account for non-conservative forces acting upon the system, such as damping and control inputs. The networks are trained to learn from simulated data of a second-order ordinary differential equation of an AUV. The trained model is evaluated by integrating paths from different initial states and comparing them to the true dynamics. The results yielded a model capable of predicting the output acceleration of the state space model but struggled in learning the direction of the forward movement with time.I den här uppsatsen utforskas Lagrangianska Neurala Nätverk (LNN) för systemidentifiering av Autonoma Undervattensfordon (AUV) med 6 frihetsgrader. En av de största utmaningarna med AUV är deras begränsningar när det kommer till trådlös kommunikation och navigering under vatten. Ett krav för att ha fungerande AUV är deras förmåga att navigera och utföra uppdrag under okända undervattensförhållanden med begränsad och brusig sensordata. Dessutom är ett kritiskt krav för lokalisering och adaptiv reglerteknik att ha noggranna modeller av systemets olinjära dynamik, samtidigt som den dynamiska miljön i havet tas i beaktande. De flesta sådana modeller tar inte i beaktande sensordata för att reglera dess parameterar. Insamling av sådan data för AUVer är besvärligt, men nödvändigt för att skapa större flexibilitet hos modellens parametrar. Trots de senaste genombrotten inom djupinlärning är traditionella metoder av systemidentifiering dominanta än idag för AUV. Det är av dessa anledningar som vi i denna uppsats strävar efter en datadriven metod, där vi förankrar lagar från fysik under inlärningen av systemets state-space modell. Mer specifikt utforskar vi LNN för ett system med högre dimension. Vidare expanderar vi även LNN till att även ta ickekonservativa krafter som verkar på systemet i beaktande, såsom dämpning och styrsignaler. Nätverket tränas att lära sig från simulerad data från en andra ordningens differentialekvation som beskriver en AUV. Den tränade modellen utvärderas genom att iterativt integrera fram dess rörelse från olika initialstillstånd, vilket jämförs med den korrekta modellen. Resultaten visade en modell som till viss del var kapabel till att förutspå korrekt acceleration, med begränsad framgång i att lära sig korrekt rörelseriktning framåt i tiden.
3
Jing, Li Ph D. Massachusetts Institute of Technology. "Physical symmetry enhanced neural networks." Thesis, Massachusetts Institute of Technology, 2020. https://hdl.handle.net/1721.1/128294.
Full textAbstract:
This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.Thesis: Ph. D., Massachusetts Institute of Technology, Department of Physics, February, 2020
Cataloged from student-submitted PDF version of thesis
Includes bibliographical references (pages 91-99).
Artificial Intelligence (AI), widely considered "the fourth industrial revolution", has shown its potential to fundamentally change our world. Today's AI technique relies on neural networks. In this thesis, we propose several physical symmetry enhanced neural network models. We first developed unitary recurrent neural networks (RNNs) that solve gradient vanishing and gradient explosion problems. We propose an efficient parametrization method that requires [sigma] (1) complexity per parameter. Our unitary RNN model has shown optimal long-term memory ability. Next, we combine the above model with a gated mechanism. This model outperform popular recurrent neural networks like long short-term memory (LSTMs) and gated recurrent units (GRUs) in many sequential tasks. In the third part, we develop a convolutional neural network architecture that achieves logarithmic scale complexity using symmetry breaking concepts. We demonstrate that our model has superior performance on small image classification tasks. In the last part, we propose a general method to extend convolutional neural networks' inductive bias and embed other types of symmetries. We show that this method improves prediction performance on lens-distorted image
by Li Jing.
Ph. D.
Ph.D. Massachusetts Institute of Technology, Department of Physics
4
Sutherland, Connie. "Spatio-temporal feedback in stochastic neural networks." Thesis, University of Ottawa (Canada), 2007. http://hdl.handle.net/10393/27559.
Full textAbstract:
The mechanisms by which groups of neurons interact is an important facet to understanding how the brain functions. Here we study stochastic neural networks with delayed feedback. The first part of our study looks at how feedback and noise affect the mean firing rate of the network. Secondly we look at how the spatial profile of the feedback affects the behavior of the network.
Our numerical and theoretical results show that negative (inhibitory) feedback linearizes the frequency vs input current (f-I) curve via the divisive gain effect it has on the network. The interaction of the inhibitory feedback and the input bias is what produces the divisive decrease in the slope (known as the gain) of the f-I curve. Our work predicts that an increase in noise is required along with increase in inhibitory feedback to attain a divisive and subtractive shift of the gain as seen in experiments [1].
Our results also show that, although the spatial profile of the feedback does not effect the mean activity of the network, it does influence the overall dynamics of the network. Local feedback generates a network oscillation, which is more robust against disruption by noise or uncorrelated input or network heterogeneity, than that for the global feedback (all-to-all coupling) case. For example uncorrelated input completely disrupts the network oscillation generated by global feedback, but only diminishes the network oscillation due to local feedback. This is characterized by 1st and 2nd order spike train statistics. Further, our theory agrees well with numerical simulations of network dynamics.
5
Squadrani, Lorenzo. "Deep neural networks and thermodynamics." Bachelor's thesis, Alma Mater Studiorum - Università di Bologna, 2020.
Find full textAbstract:
Deep learning is the most effective and used approach to artificial intelligence, and yet it is far from being properly understood. The understanding of it is the way to go to further improve its effectiveness and in the best case to gain some understanding of the "natural" intelligence. We attempt a step in this direction with the aim of physics. We describe a convolutional neural network for image classification (trained on CIFAR-10) within the descriptive framework of Thermodynamics. In particular we define and study the temperature of each component of the network. Our results provides a new point of view on deep learning models, which may be a starting point towards a better understanding of artificial intelligence.
6
Gyawali, Gaurav. "Solving Atomic Wave Functions Using Artificial Neural Networks." ScholarWorks@UNO, 2018. https://scholarworks.uno.edu/honors_theses/104.
Full textAbstract:
Carleo and Troyer [3] have recently pointed out the possibility of solving quantum many-body problems by using Artificial Neural Networks (ANN). Their work is based on minimizing a variational wave function to obtain the ground states for various spin-dependent systems. This work is primarily focused on developing efficient method using ANN to solve the ground state wave function for atomic systems. We have developed a theoretical groundwork to represent the wave function of a many-electron atom by using artificial neural network while still preserving its antisymmetric property. By using the Metropolis algorithm, Variational Monte Carlo (VMC), and Stochastic Reconfiguration (SR) methods for minimization, we were able to obtain a highly accurate ground state wave function for the He atom. To verify our optimization algorithm, we reproduced the results for the ground state of a three dimensional Simple Harmonic Oscillator (SHO) given by Teng [18].
7
Düring, Alexander. "Temporal aspects of spin-glass neural networks." Thesis, University of Oxford, 1998. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.325892.
Full text
8
Wu, Dawen. "Solving Some Nonlinear Optimization Problems with Deep Learning." Electronic Thesis or Diss., université Paris-Saclay, 2023. http://www.theses.fr/2023UPASG083.
Full textAbstract:
Cette thèse considère quatre types de problèmes d'optimisation non linéaire, à savoir les jeux de bimatrice, les équations de projection non linéaire (NPEs), les problèmes d'optimisation convexe non lisse (NCOPs) et les jeux à contraintes stochastiques (CCGs). Ces quatre classes de problèmes d'optimisation non linéaire trouvent de nombreuses applications dans divers domaines tels que l'ingénierie, l'informatique, l'économie et la finance. Notre objectif est d'introduire des algorithmes basés sur l'apprentissage profond pour calculer efficacement les solutions optimales de ces problèmes d'optimisation non linéaire.Pour les jeux de bimatrice, nous utilisons des réseaux neuronaux convolutionnels (CNNs) pour calculer les équilibres de Nash. Plus précisément, nous concevons une architecture de CNN où l'entrée est un jeu de bimatrice et la sortie est l'équilibre de Nash prédit pour le jeu. Nous générons un ensemble de jeux de bimatrice suivant une distribution de probabilité donnée et utilisons l'algorithme de Lemke-Howson pour trouver leurs véritables équilibres de Nash, constituant ainsi un ensemble d'entraînement. Le CNN proposé est formé sur cet ensemble de données pour améliorer sa précision. Une fois l'apprentissage terminée, le CNN est capable de prédire les équilibres de Nash pour des jeux de bimatrice inédits. Les résultats expérimentaux démontrent l'efficacité computationnelle exceptionnelle de notre approche basée sur CNN, au détriment de la précision.Pour les NPEs, NCOPs et CCGs, qui sont des problèmes d'optimisation plus complexes, ils ne peuvent pas être directement introduits dans les réseaux neuronaux. Par conséquent, nous avons recours à des outils avancés, à savoir l'optimisation neurodynamique et les réseaux neuronaux informés par la physique (PINNs), pour résoudre ces problèmes. Plus précisément, nous utilisons d'abord une approche neurodynamique pour modéliser un problème d'optimisation non linéaire sous forme de système d'équations différentielles ordinaires (ODEs). Ensuite, nous utilisons un modèle basé sur PINN pour résoudre le système d'ODE résultant, où l'état final du modèle représente la solution prédite au problème d'optimisation initial. Le réseau neuronal est formé pour résoudre le système d'ODE, résolvant ainsi le problème d'optimisation initial. Une contribution clé de notre méthode proposée réside dans la transformation d'un problème d'optimisation non linéaire en un problème d'entraînement de réseau neuronal. En conséquence, nous pouvons maintenant résoudre des problèmes d'optimisation non linéaire en utilisant uniquement PyTorch, sans compter sur des solveurs d'optimisation convexe classiques tels que CVXPY, CPLEX ou GurobiThis thesis considers four types of nonlinear optimization problems, namely bimatrix games, nonlinear projection equations (NPEs), nonsmooth convex optimization problems (NCOPs), and chance-constrained games (CCGs).These four classes of nonlinear optimization problems find extensive applications in various domains such as engineering, computer science, economics, and finance.We aim to introduce deep learning-based algorithms to efficiently compute the optimal solutions for these nonlinear optimization problems.For bimatrix games, we use Convolutional Neural Networks (CNNs) to compute Nash equilibria.Specifically, we design a CNN architecture where the input is a bimatrix game and the output is the predicted Nash equilibrium for the game.We generate a set of bimatrix games by a given probability distribution and use the Lemke-Howson algorithm to find their true Nash equilibria, thereby constructing a training dataset.The proposed CNN is trained on this dataset to improve its accuracy. Upon completion of training, the CNN is capable of predicting Nash equilibria for unseen bimatrix games.Experimental results demonstrate the exceptional computational efficiency of our CNN-based approach, at the cost of sacrificing some accuracy.For NPEs, NCOPs, and CCGs, which are more complex optimization problems, they cannot be directly fed into neural networks.Therefore, we resort to advanced tools, namely neurodynamic optimization and Physics-Informed Neural Networks (PINNs), for solving these problems.Specifically, we first use a neurodynamic approach to model a nonlinear optimization problem as a system of Ordinary Differential Equations (ODEs).Then, we utilize a PINN-based model to solve the resulting ODE system, where the end state of the model represents the predicted solution to the original optimization problem.The neural network is trained toward solving the ODE system, thereby solving the original optimization problem.A key contribution of our proposed method lies in transforming a nonlinear optimization problem into a neural network training problem.As a result, we can now solve nonlinear optimization problems using only PyTorch, without relying on classical convex optimization solvers such as CVXPY, CPLEX, or Gurobi
9
Tolley, Emma Elizabeth. "Monte Carlo event reconstruction implemented with artificial neural networks." Thesis, Massachusetts Institute of Technology, 2011. http://hdl.handle.net/1721.1/65535.
Full textAbstract:
Thesis (S.B.)--Massachusetts Institute of Technology, Dept. of Physics, 2011.Cataloged from PDF version of thesis.
Includes bibliographical references (p. 41).
I implemented event reconstruction of a Monte Carlo simulation using neural networks. The OLYMPUS Collaboration is using a Monte Carlo simulation of the OLYMPUS particle detector to evaluate systematics and reconstruct events. This simulation registers the passage of particles as 'hits' in the detector elements, which can be used to determine event parameters such as momentum and direction. However, these hits are often obscured by noise. Using Geant4 and ROOT, I wrote a program that uses artificial neural networks to separate track hits from noise and reconstruct event parameters. The classification network successfully discriminates between track hits and noise for 97.48% of events. The reconstruction networks determine the various event parameters to within 2-3%.
by Emma Elizabeth Tolley.
S.B.
10
Doriat, Aurélien. "Caractérisation des couplages aéro-thermo-mécaniques lors d’un vieillissement par thermo-oxydation de composites à matrice polymère soumis à un écoulement rapide et chauffé." Electronic Thesis or Diss., Chasseneuil-du-Poitou, Ecole nationale supérieure de mécanique et d'aérotechnique, 2024. http://www.theses.fr/2024ESMA0018.
Full textAbstract:
Les matériaux composites à matrice organique renforcés de fibres de carbone (CFRP) sont largement utilisés dans les structures aéronautiques froides. Dans les applications de moteurs aéronautiques, comme les aubes de FAN, ces matériaux peuvent être soumis à des conditions environnementales particulièrement sévères, avec des températures pouvant atteindre 120°C et une vitesse d’écoulement proche de Mach 1.Il est bien établi que les polymères époxy sont sujets à des phénomènes de thermo-oxydation lorsqu’ils sont exposés à des températures élevées. Ce phénomène implique la diffusion et la réaction de l’oxygène au sein du polymère, entraînant des changements de couleur, une antiplasticisation du matériau, une fragilisation. Jusqu’à présent, les essais de vieillissement ont été principalement effectués dans des fours à air statique, fournissant une compréhension détaillée du phénomène dans ces conditions. Cependant, l’impact de l’écoulement d’air sur la thermo-oxydation reste à explorer. Cette étude vise ainsi à approfondir la compréhension du couplage entre l’écoulement d’air et la dégradation du matériau par thermo-oxydation. Des échantillons ont été vieillis dans un four sous air à pression atmosphérique et dans la soufflerie BATH, adaptée pour ces essais et capable de générer un écoulement d’air à plus de 150 ◦C et Mach 1, reproduisant ainsi les conditions d’usage les plus sévère rencontrées dans des moteurs d’avion. Cette comparaison entre essai en four et soufflerie a montré une accélération du vieillissement en soufflerie. Pour obtenir ce résultat, une technique expérimentale basée sur le changement de couleur induit par l’oxydation a été développée et utilisée. Cette technique a été validée avec des essais d’indentation. Avec cette meilleure compréhension de l’accélération du vieillissement, un modèle couplé entre l’écoulement, la chimie de l’oxydation et les changements de propriétés mécaniques a été mis en place afin de mieux comprendre les mécanismes à l’interface. Cette modélisation comprends trois étapes. Les champs de pression et de température à la surface de l’échantillon ont été calculés par simulation fluide moyennée (RANS). Puis, un modèle mécanistique a été utilisé décrivant les réactions chimiques lors de l’oxydation. Enfin, sur la base des mesures de couleur, un réseau de neurones informé par la physique (PINN) a été mis en place pour coupler les quantités chimiques aux propriétés mécaniquesCarbon fiber-reinforced polymer matrix composites (CFRP) are widely used in cold aeronautical structures. In aeronautical engine applications, such as fan blades, these materials can be subjected to particularly severe environmental conditions, with temperatures reaching up to 120 ◦C and airflow speeds close to Mach 1. It is well established that epoxy polymers are prone to thermo-oxidation phenomena when exposed to high temperatures.This phenomenon involves the diffusion and reaction of oxygen within the polymer, leading to color changes, antiplasticization of the material, and embrittlement. Until now, aging tests have been mainly conducted in static air ovens, providing a detailed understanding of the phenomenon under these conditions. However, the impact of airflow on thermo-oxidation remains to be explored.This study thus aims to deepen the understanding of the coupling between airflow and material degradation due to thermo-oxidation.Samples were aged in an oven under air at atmospheric pressure and in the BATH wind tunnel, adapted for these tests and capable of generating an airflow at over 150 ◦C and Mach 1, thereby reproducing the most severe usage conditions encountered in aircraft engines. This comparison between oven and wind tunnel tests showed an acceleration of aging in the wind tunnel. To achieve this result, an experimental technique based on the color change induced by oxidation was developed and used. This technique was validated with indentation tests. With this improved understanding of the accelerated aging, a coupled model between the airflow, oxidation chemistry, and changes in mechanical properties was established to better understand the interfacial mechanisms. This modeling comprises three steps. The pressure and temperature fields at the sample surface were calculated using Reynolds-Averaged Navier-Stokes (RANS) fluid simulations. Then, a mechanistic model was used to describe the chemical reactions during oxidation. Finally, based on thecolor measurements, a physics-informed neural network (PINN) was implemented to couple the chemical quantities to the mechanical properties
11
Penney, Richard William. "The statistical mechanics of neural networks and spin glasses." Thesis, University of Oxford, 1994. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.239333.
Full text
12
Andersson, Mikael. "Gamma-ray racking using graph neural networks." Thesis, KTH, Fysik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-298610.
Full textAbstract:
While there are existing methods of gamma ray-track reconstruction in specialized detectors such as AGATA, including backtracking and clustering, it is naturally of interest to diversify the portfolio of available tools to provide us viable alternatives. In this study some possibilities found in the field of machine learning were investigated, more specifically within the field of graph neural networks. In this project there was attempt to reconstruct gamma tracks in a germanium solid using data simulated in Geant4. The data consists of photon energies below the pair production limit and so we are limited to the processes of photoelectric absorption and Compton scattering. The author turned to the field of graph networks to utilize its edge and node structure for data of such variable input size as found in this task. A graph neural network (GNN) was implemented and trained on a variety of gamma multiplicities and energies and was subsequently tested in terms of various accuracy parameters and generated energy spectra. In the end the best result involved an edge classifier trained on a large dataset containing a 10^6 tracks bundled together into separate events to be resolved. The network was capable of recalling up to 95 percent of the connective edges for the selected threshold in the infinite resolution case with a peak-to-total ratio of 68 percent for a set of packed data with a model trained on simulated data including realistic uncertainties in both position and energy.Trots att det existerar en mängd metoder för rekonstruktion av spår i specialiserade detektorer som AGATA är det av naturligt intresse att diversifiera och undersöka nya verktyg för uppgiften. I denna studie undersöktes några möjligheter inom maskininlärning, närmare bestämt inom området neurala grafnätverk. Under projektets gång simulerades data av fotoner i en ihålig, sfärisk geometri av germanium i Geant4. Den simulerade datan är begränsad till energier under parproduktion så datan består av reaktioner genom den fotoelektriska effekten och comptonspridning. Den variabla storleken på denna data och dess spridning i detektorns geometri lämpar sig för ett grafformat med nod och länkstruktur. Ett neuralt grafnätverk (GNN) implementerades och tränades på data med gamma av variabel multiplicitet och energi och evaluerades på ett urval prestandaparametrar och dess förmåga att generera ett användbart spektra. Slutresultatet involverade en länkklassificerings modell tränat på data med 10^6 spår sammanslagna till händelser. Nätverket återkallade 95 procent av positiva länkar för ett val av tröskelvärde i fallet med oändlig upplösning med ett peak-to-total på 68 procent för packad data behandlad med osäkerhet i energi och position motsvarande fallet med begränsad upplösning.
13
Andersson, Mikael. "Gamma-ray tracking using graph neural networks." Thesis, KTH, Fysik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-298610.
Full textAbstract:
While there are existing methods of gamma ray-track reconstruction in specialized detectors such as AGATA, including backtracking and clustering, it is naturally of interest to diversify the portfolio of available tools to provide us viable alternatives. In this study some possibilities found in the field of machine learning were investigated, more specifically within the field of graph neural networks. In this project there was attempt to reconstruct gamma tracks in a germanium solid using data simulated in Geant4. The data consists of photon energies below the pair production limit and so we are limited to the processes of photoelectric absorption and Compton scattering. The author turned to the field of graph networks to utilize its edge and node structure for data of such variable input size as found in this task. A graph neural network (GNN) was implemented and trained on a variety of gamma multiplicities and energies and was subsequently tested in terms of various accuracy parameters and generated energy spectra. In the end the best result involved an edge classifier trained on a large dataset containing a 10^6 tracks bundled together into separate events to be resolved. The network was capable of recalling up to 95 percent of the connective edges for the selected threshold in the infinite resolution case with a peak-to-total ratio of 68 percent for a set of packed data with a model trained on simulated data including realistic uncertainties in both position and energy.Trots att det existerar en mängd metoder för rekonstruktion av spår i specialiserade detektorer som AGATA är det av naturligt intresse att diversifiera och undersöka nya verktyg för uppgiften. I denna studie undersöktes några möjligheter inom maskininlärning, närmare bestämt inom området neurala grafnätverk. Under projektets gång simulerades data av fotoner i en ihålig, sfärisk geometri av germanium i Geant4. Den simulerade datan är begränsad till energier under parproduktion så datan består av reaktioner genom den fotoelektriska effekten och comptonspridning. Den variabla storleken på denna data och dess spridning i detektorns geometri lämpar sig för ett grafformat med nod och länkstruktur. Ett neuralt grafnätverk (GNN) implementerades och tränades på data med gamma av variabel multiplicitet och energi och evaluerades på ett urval prestandaparametrar och dess förmåga att generera ett användbart spektra. Slutresultatet involverade en länkklassificerings modell tränat på data med 10^6 spår sammanslagna till händelser. Nätverket återkallade 95 procent av positiva länkar för ett val av tröskelvärde i fallet med oändlig upplösning med ett peak-to-total på 68 procent för packad data behandlad med osäkerhet i energi och position motsvarande fallet med begränsad upplösning.
14
Cardoso, Mário. "Study of pattern recognition of particle tracks with neural networks." Thesis, Uppsala universitet, Högenergifysik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-454374.
Full textAbstract:
In this project we study the use of neural networks as a tool for particle track pattern recognition with the possibility of its implementation in the Trigger system at the ATLAS experiment [1]. By using a method named Hough transform we created a Convolutional Neural Network (CNN) that is able to train on the transformed images of muons merged with minimum bias. We give an overview of how the CNN works and compare the results from the CNN with the old cut based method. We believe to have managed to find an alternative to the previously used algorithm, that is faster and more efficient.
15
Skirlo, Scott Alexander. "Photonics for technology : circuits, chip-scale LIDAR, and optical neural networks." Thesis, Massachusetts Institute of Technology, 2017. http://hdl.handle.net/1721.1/112519.
Full textAbstract:
Thesis: Ph. D., Massachusetts Institute of Technology, Department of Physics, 2017.Cataloged from PDF version of thesis.
Includes bibliographical references (pages 163-175).
This thesis focuses on a wide range of contemporary topics in modern electromagnetics and technology including topologically protected one-way modes, integrated photonic LIDAR, and optical neural networks. First, we numerically investigate large Chern numbers in photonic crystals and explore their origin from simultaneously gapping multiple band degeneracies. Following this, we perform microwave transmission measurements in the bulk and at the edge of ferrimagnetic photonic crystals. Bandgaps with large Chern numbers of 2, 3, and 4 are present in the experimental results 'which show excellent agreement with theory. We measure the mode profiles and Fourier transform them to produce dispersion relations of the edge modes, whose number and direction match our Chern number calculations. We use these waveguides to realize reflectionless power splitters and outline their application to general one-way circuits. Next we create a new chip-scale LIDAR architecture in analogy to planar RF lenses. Instead of relying upon many continuously tuned thermal phase shifters to implement nonmechanical beam steering, we use aplanatic lenses excited in their focal plane feeding ID gratings to generate discrete beams. We design devices which support up to 128 resolvable points in-plane and 80 resolvable points out-of-plane, which are currently being fabricated and tested. These devices have many advantages over conventional optical phased arrays including greatly increased optical output power and decreased electrical power for in-plane beamforming. Finally we explore a new approach for implementing convolutional neural networks through an integrated photonics circuit consisting of Mach-Zehnder Interferometers, optical delay lines, and optical nonlinearity units. This new platform, should be able to perform the order of a thousand inferences per second, at [mu]J power levels per inference, with the nearest state of the art ASIC and GPU competitors operating 30 times slower and requiring three orders of magnitude more power.
by Scott Alexander Skirlo.
Ph. D.
16
Macpherson, Keith P. "Prediction of solar and geomagnetic activity using artificial neural networks." Thesis, University of Glasgow, 1994. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.296375.
Full text
17
Davila, Carlos Antonio. "Image super-resolution performance of multilayer feedforward neural networks." Diss., The University of Arizona, 1999. http://hdl.handle.net/10150/284549.
Full textAbstract:
Super-resolution is the process by which the bandwidth of a diffraction-limited spectrum is extended beyond the optical passband. Many algorithms exist which are capable of super-resolution; however most are iterative methods, which are ill-suited for real-time operation. One approach that has been virtually ignored in super-resolution research is the neural network approach. The Hopfield network has been a popular choice in image restoration applications, however it is also an iterative approach. We consider the feedforward architecture known as a Multilayer Perceptron (MLP), and present results on simulated binary and greyscale images blurred by a diffraction-limited OTF and sampled at the Nyquist rate. To avoid aliasing, the network performs as a nonlinear spatial interpolator while simultaneously extrapolating in the frequency domain. Additionally, a novel use of vector quantization for the generation of training data sets is presented. This is accomplished by training a nonlinear vector quantizer (NLIVQ), whose codebooks are subsequently used in the supervised training of the MLP network using Back-Propagation. The network shows good regularization in the presence of noise.
18
Hodges, Jonathan Lee. "Predicting Large Domain Multi-Physics Fire Behavior Using Artificial Neural Networks." Diss., Virginia Tech, 2018. http://hdl.handle.net/10919/86364.
Full textAbstract:
Fire dynamics is a complex process involving multi-mode heat transfer, reacting fluid flow, and the reaction of combustible materials. High-fidelity predictions of fire behavior using computational fluid dynamics (CFD) models come at a significant computational cost where simulation times are often measured in hours, days, or even weeks. A new simulation method is to use a machine learning approach which uses artificial neural networks (ANNs) to represent underlying connections between data to make predictions of new inputs. The field of image analysis has seen significant advancements in ANN performance by using feature based layers in the network architecture. Inspired by these advancements, a generalized procedure to design ANNs to make spatially resolved predictions in multi-physics applications is presented and applied to different fire applications. A deep convolutional inverse graphics network (DCIGN) was developed to predict the two-dimensional spatially resolved spread of a wildland fire. The network uses an image stack corresponding to the spatially resolved landscape, weather, and current fire perimeter (which can be obtained from measurements) to predict the fire perimeter six hours in the future. A transpose convolutional neural network (TCNN) was developed to predict the spatially resolved thermal flow field in a compartment fire from coarse zone fire model predictions. The network uses thirty-five parameters describing the geometry of the room and the ventilation conditions to predict the full-field temperature and velocity throughout the room. The data for use in training and testing both networks was generated using high-fidelity CFD fire simulations. Overall, the ANN predictions in each network agree with simulation predictions for validation scenarios. The computational time to evaluate the ANNs is 10,000x faster than the high-fidelity fire simulations. This work represents a first step in developing super real-time full-field fire predictions for different applications.Ph. D.
The National Fire Protection Agency estimates the total cost of fire in the United States at $300 billion annually. In 2017 alone, there were 3,400 civilian fire fatalities, 14,670 civilian fire injuries, and an estimated $23 billion direct property loss in the United States. Large scale fires in the wildland urban interface (WUI) and in large buildings still represent a significant hazard to life, property, and the environment. Researchers and fire safety engineers often use computer simulations to predict the behavior of a fire to assist in reducing the hazard of fire. Unfortunately, typical simulations of fire scenarios may take hours, days, or even weeks to run which limits their use to small areas or sections of buildings. A new method is to use a machine learning approach which uses artificial neural networks (ANNs) to represent underlying connections between data to make new predictions of fire behavior. Inspired by advancements in the field of image processing, this research developed a procedure to use machine learning to make rapid high resolution predictions of fire behavior. An ANN was developed to predict the perimeter of a wildland fire six hours in the future based on a set of images corresponding to the landscape, weather, and current fire perimeter, all of which can be obtained directly from measurements (US Geological Survey, Automated Surface Observation System, and satellites). In addition, an ANN was developed to predict high-resolution temperature and velocity fields within a floor of a building based on predictions from a coarse model. The data for use in training and testing these networks was generated using high-resolution fire simulations. Overall, the network predictions agree well with simulation predictions for new scenarios. In addition, the time to run the model is 10,000x faster than the typical simulations. The work presented herein represents a first step in developing high resolution computer simulations for different fire scenarios that run very quickly.
19
Ajay, Anurag. "Augmenting physics simulators with neural networks for model learning and control." Thesis, Massachusetts Institute of Technology, 2019. https://hdl.handle.net/1721.1/122747.
Full textAbstract:
Thesis: S.M., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2019Cataloged from PDF version of thesis.
Includes bibliographical references (pages 77-81).
Physics simulators play an important role in robot state estimation, planning and control; however, many real-world control problems involve complex contact dynamics that cannot be characterized analytically. Therefore, most physics simulators employ approximations that lead to a loss in precision. We propose a hybrid dynamics model, combining a deterministic physical simulator with a stochastic neural network for dynamics modeling as it provides us with expressiveness, efficiency, and generalizability simultaneously. To demonstrate this, we compare our hybrid model to both purely analytical models and purely learned models. We then show that our model is able to characterize the complex distribution of object trajectories and compare it with existing methods. We further build in object based representation into the neural network so that our hybrid model can generalize across number of objects. Finally, we use our hybrid model to complete complex control tasks in simulation and on a real robot and show that our model generalizes to novel environments with varying object shapes and materials.
by Anurag Ajay.
S.M.
S.M. Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science
20
Sánchez, Carlos Andrés. "Measurement of the Top Quark Mass with Neural Networks." The Ohio State University, 2002. http://rave.ohiolink.edu/etdc/view?acc_num=osu1039046089.
Full text
21
Roy, Chandan. "An Informed System Development Approach to Tropical Cyclone Track and Intensity Forecasting." Doctoral thesis, Linköpings universitet, Interaktiva och kognitiva system, 2016. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-123198.
Full textAbstract:
Introduction: Tropical Cyclones (TCs) inflict considerable damage to life and property every year. A major problem is that residents often hesitate to follow evacuation orders when the early warning messages are perceived as inaccurate or uninformative. The root problem is that providing accurate early forecasts can be difficult, especially in countries with less economic and technical means. Aim: The aim of the thesis is to investigate how cyclone early warning systems can be technically improved. This means, first, identifying problems associated with the current cyclone early warning systems, and second, investigating if biologically based Artificial Neural Networks (ANNs) are feasible to solve some of the identified problems. Method: First, for evaluating the efficiency of cyclone early warning systems, Bangladesh was selected as study area, where a questionnaire survey and an in-depth interview were administered. Second, a review of currently operational TC track forecasting techniques was conducted to gain a better understanding of various techniques’ prediction performance, data requirements, and computational resource requirements. Third, a technique using biologically based ANNs was developed to produce TC track and intensity forecasts. Systematic testing was used to find optimal values for simulation parameters, such as feature-detector receptive field size, the mixture of unsupervised and supervised learning, and learning rate schedule. Five types of 2D data were used for training. The networks were tested on two types of novel data, to assess their generalization performance. Results: A major problem that is identified in the thesis is that the meteorologists at the Bangladesh Meteorological Department are currently not capable of providing accurate TC forecasts. This is an important contributing factor to residents’ reluctance to evacuate. To address this issue, an ANN-based TC track and intensity forecasting technique was developed that can produce early and accurate forecasts, uses freely available satellite images, and does not require extensive computational resources to run. Bidirectional connections, combined supervised and unsupervised learning, and a deep hierarchical structure assists the parallel extraction of useful features from five types of 2D data. The trained networks were tested on two types of novel data: First, tests were performed with novel data covering the end of the lifecycle of trained cyclones; for these test data, the forecasts produced by the networks were correct in 91-100% of the cases. Second, the networks were tested with data of a novel TC; in this case, the networks performed with between 30% and 45% accuracy (for intensity forecasts). Conclusions: The ANN technique developed in this thesis could, with further extensions and up-scaling, using additional types of input images of a greater number of TCs, improve the efficiency of cyclone early warning systems in countries with less economic and technical means. The thesis work also creates opportunities for further research, where biologically based ANNs can be employed for general-purpose weather forecasting, as well as for forecasting other severe weather phenomena, such as thunderstorms.
22
Rosten, David Paul 1967. "Automatic design of a decision tree classifier employing neural networks." Thesis, The University of Arizona, 1991. http://hdl.handle.net/10150/277881.
Full textAbstract:
Pattern recognition problems involve two main issues: feature formulation and classifier design. This thesis is concerned with the latter. Numerous algorithms for the design of pattern recognition systems have been published, and the algorithm detailed herein is a new approach--specific to the design of decision tree classifiers. It involves a top-down strategy, optimizing the root node decision and then subsequently its children. To assess various pattern space partitions, the Tie statistical distance measure quantified the separability of potential cluster groupings. Additionally, a separate neural network was employed at each of the tree decision nodes. Results from the application of this methodology to the regional labeling of panchromatic images suggest it is a suitable approach.
23
Stevenson, King Douglas Beverley. "Robust hardware elements for weightless artificial neural networks." Thesis, University of Central Lancashire, 2000. http://clok.uclan.ac.uk/1884/.
Full textAbstract:
This thesis investigates novel robust hardware elements for weightless artificial neural systems with a bias towards high integrity avionics applications. The author initially reviews the building blocks of physiological neural systems and then chronologically describes the development of weightless artificial neural systems. Several new design methodologies for the implementation of robust binary sum-and-threshold neurons are presented. The new techniques do not rely on weighted binary counters or registered arithmetic units for their operation making them less susceptible to transient single event upsets. They employ Boolean, weightless binary, asynchronous elements throughout thus increasing robustness in the presence of impulsive noise. Hierarchies formed from these neural elements are studied and a weightless probabilisitic activation function proposed for non-deterministic applications. Neuroram, an auto-associative memory created using these weightless neurons is described and analysed. The signal-to-noise ratio characteristics are compared with the traditional Hamming distance metric. This led to the proposal that neuroram can form a threshold logic based digital signal filter. Two weightless autoassociative memory based neuro-filters are presented and their filtration properties studied and compared with a traditional median filter. Eachn novel architecture was emulated using weightless numericM ATLAB code prior to schematic design and functional simulation. Several neural elements were implemented and validated using FPGA technology. A preliminary robustness evaluation was performed. The large scale particle accelerator at the Theodor Svedberg Laboratory at the University of Uppsala, Sweden, was used to generate transienut psetsin an FPGA performing a weightless binary neural function. One paper,two letters and five international patents have been published during the course of this research. The author has significantly contributed to the field of weightless artificial neural systems in high integrity hardware applications.
24
Hedström, Lucas. "Classifying the rotation of bacteria using neural networks." Thesis, Umeå universitet, Institutionen för fysik, 2019. http://urn.kb.se/resolve?urn=urn:nbn:se:umu:diva-160518.
Full textAbstract:
Bacteria can quickly spread throughout the human body, making certain diseases hard or impossible to cure. In order to understand how the bacteria can initiate and develop into an infection, microfluidic chambers in a lab environment are used as a template of how bacteria reacts to different types of flows. However, accurately tracking the movement of bacteria is a difficult task, where small objects has to be captured with a high resolution and digitally analysed with computationally heavy methods. Popular imaging methods utilise digital holographic microscopy, where three-dimensional movement is captured in two-dimensional images by numerical reconstruction of the diffraction of light. Since numerical reconstructions become computationally heavy when a good accuracy is required, this master's thesis work focus on evaluating the possibility of using convolutional neural networks to quickly and accurately determine the spatial properties of bacteria. By thorough testing and analysis of state of the art and old networks a new network design is presented, designed to eliminate as many imaging issues as possible. We found that there are certain network design choices that help with reducing the overall error of the system, and with a well chosen training set with sensible augmentations, some networks were able to reach a 60% classification accuracy when determining the vertical rotation of the bacteria. Unfortunately, due to the lack of experimental data where the ground-truth is known, not much experimental testing could be performed. However, a few tests showed that images of high quality could be classified within the expected range of vertical rotation.Bakterier kan snabbt sprida sig genom människokroppen, vilket försvårar starkt möjligheterna att kurera vissa sjukdomar. För att få en inblick i hur bakterier kan initiera och utvecklas till en infektion används som mall laborativa uppställningar med vätskekanaler i mikroskala när man söker förstå hur bakterier reagerar på olika typer av flöden. Att spåra dessa rörelser med god säkerhet är dock en utmaning, då man experimentellt söker fånga små skalor med hög upplösning, som sedan ska analyseras med datorintensiva metoder. Populära avbildningsmetoder använder sig utav digital holografisk mikroskopi, där tredimensionella rörelser kan fångas med hjälp av tvådimensionella bilder genom att numeriskt återskapa ljusets brytningsmönster mot objekten. Eftersom dessa metoder blir beräkningstunga när god säkerhet krävs så utforskar detta examensarbete möjligheterna att utnyttja faltningsnätverk för att snabbt och säkert bestämma vertikalrotationen hos bakterier avbildade med holografi. Genom nogranna tester av moderna samt äldre nätverk så presenteras en ny nätverksdesign, utvecklad i mål med att eliminera så många avbildningsproblem som möjligt. Vi fann att vissa designval vid nätverksutvecklingen kan hjälpa med att reducera klassificeringsfelen givet vårt system, och med en väl utvald träningsmängd med lämpliga justeringar så lyckades vi nå en klassificeringssäkerhet på över 60% med vissa nätverk. På grund av bristande experimentellt data där de riktiga värdena är kända så har ingen utförlig experimentell analys utförts, men några tester på experimentella bilder i god kvalité har visats ge resultat som tyder på en korrekt analys inom den förväntade vertikalrotationen.
25
Rigon, Luca. "Development of an intelligent trigger system based on deep neural networks." Thesis, Uppsala universitet, Högenergifysik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-446864.
Full text
26
Ciobanu, Cătălin Irinel. "A neural networks search for single top quark production in CDF Run I Data /." The Ohio State University, 2002. http://rave.ohiolink.edu/etdc/view?acc_num=osu1486461246815552.
Full text
27
Mignacco, Francesca. "Statistical physics insights on the dynamics and generalisation of artificial neural networks." Thesis, université Paris-Saclay, 2022. http://www.theses.fr/2022UPASP074.
Full textAbstract:
L'apprentissage machine est une technologie désormais omniprésente dans notre quotidien. Toutefois, ce domaine reste encore largement empirique et ses enjeux scientifiques manquent d'une compréhension théorique profonde. Cette thèse se penche vers la découverte des mécanismes sous-tendant l'apprentissage dans les réseaux de neurones artificiels à travers le prisme de la physique statistique. Dans une première partie, nous nous intéressons aux propriétés statiques des problèmes d'apprentissage, que nous introduisons au chapitre 1.1. Dans le chapitre 1.2, nous considérons la classification d'un mélange binaire de nuages gaussiens et nous dérivons des expressions rigoureuses pour les erreurs en dimension infinie, que nous appliquons pour éclairer le rôle des différents paramètres du problème. Dans le chapitre 1.3, nous montrons comment étendre le modèle de perceptron enseignant-étudiant pour considérer la classification multi-classes, en dérivant des expressions asymptotiques pour la performance optimale et la performance de la minimisation du risque empirique régularisé. Dans la deuxième partie, nous nous concentrons sur la dynamique de l'apprentissage, que nous introduisons dans le chapitre 2.1. Dans le chapitre 2.2, nous montrons comment décrire analytiquement la dynamique de l'algorithme du gradient stochastique à échantillonage mini-lots (mini-batch SGD) dans la classification binaire de mélanges gaussiens, en utilisant la théorie dynamique du champ moyen. Le chapitre 2.3 présente une analyse du bruit effectif introduit par SGD. Dans le chapitre 2.4, nous considérons le problème de la récupération des signes comme exemple d'optimisation hautement non convexe et montrons que la stochasticité est cruciale pour la généralisation. La conclusion de la thèse est présentée dans la troisième partieMachine learning technologies have become ubiquitous in our daily lives. However, this field still remains largely empirical and its scientific stakes lack a deep theoretical understanding.This thesis explores the mechanisms underlying learning in artificial neural networks through the prism of statistical physics. In the first part, we focus on the static properties of learning problems, that we introduce in Chapter 1.1. In Chapter 1.2, we consider the prototype classification of a binary mixture of Gaussian clusters and we derive rigorous closed-form expressions for the errors in the infinite-dimensional regime, that we apply to shed light on the role of different problem parameters. In Chapter 1.3, we show how to extend the teacher-student perceptron model to encompass multi-class classification deriving asymptotic expressions for the optimal performance and the performance of regularised empirical risk minimisation. In the second part, we turn our focus to the dynamics of learning, that we introduce in Chapter 2.1. In Chapter 2.2, we show how to track analytically the training dynamics of multi-pass stochastic gradient descent (SGD) via dynamical mean-field theory for generic non convex loss functions and Gaussian mixture data. Chapter 2.3 presents a late-time analysis of the effective noise introduced by SGD in the underparametrised and overparametrised regimes. In Chapter 2.4, we take the sign retrieval problem as a benchmark highly non-convex optimisation problem and show that stochasticity is crucial to achieve perfect generalisation. The third part of the thesis contains the conclusions and some future perspectives
28
Harris, William H. (William Hunt). "Machine learning transferable physics-based force fields using graph convolutional neural networks." Thesis, Massachusetts Institute of Technology, 2020. https://hdl.handle.net/1721.1/128979.
Full textAbstract:
Thesis: S.M., Massachusetts Institute of Technology, Department of Materials Science and Engineering, 2020Cataloged from student-submitted PDF of thesis.
Includes bibliographical references (pages 22-24).
Molecular dynamics and Monte Carlo methods allow the properties of a system to be determined from its potential energy surface (PES). In the domain of crystalline materials, the PES is needed for electronic structure calculations, critical for modeling semiconductors, optical, and energy-storage materials. While first principles techniques can be used to obtain the PES to high accuracy, their computational complexity limits applications to small systems and short timescales. In practice, the PES must be approximated using a computationally cheaper functional form. Classical force field (CFF) approaches simply define the PES as a sum over independent energy contributions. Commonly included terms include bonded (pair, angle, dihedral, etc.) and non bonded (van der Waals, Coulomb, etc.) interactions, while more recent CFFs model polarizability, reactivity, and other higher-order interactions.
Simple, physically-justified functional forms are often implemented for each energy type, but this choice - and the choice of which energy terms to include in the first place - is arbitrary and often hand-tuned on a per-system basis, severely limiting PES transferability. This flexibility has complicated the quest for a universal CFF. The simplest usable CFFs are tailored to specific classes of molecules and have few parameters, so that they can be optimally parameterized using a small amount of data; however, they suffer low transferability. Highly-parameterized neural network potentials can yield predictions that are extremely accurate for the entire training set; however, they suffer over-fitting and cannot interpolate.
We develop a tool, called AuTopology, to explore the trade-offs between complexity and generalizability in fitting CFFs; focus on simple, computationally fast functions that enforce physics-based regularization and transferability; use message-passing neural networks to featurized molecular graphs and interpolate CFF parameters across chemical space; and utilize high performance computing resources to improve the efficiency of model training and usage. A universal, fast CFF would open the door to high-throughput virtual materials screening in the pursuit of novel materials with tailored properties.
by William H. Harris.
S.M.
S.M. Massachusetts Institute of Technology, Department of Materials Science and Engineering
29
Watkin, Timothy L. H. "The theory of quenched disorder : spin glasses, neural networks and statistical inference." Thesis, University of Oxford, 1992. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.315731.
Full text
30
Reis, Elohim Fonseca dos 1984. "Criticality in neural networks = Criticalidade em redes neurais." [s.n.], 2015. http://repositorio.unicamp.br/jspui/handle/REPOSIP/276917.
Full textAbstract:
Orientadores: José Antônio Brum, Marcus Aloizio Martinez de AguiarDissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Física Gleb Wataghin
Made available in DSpace on 2018-08-29T15:40:55Z (GMT). No. of bitstreams: 1 Reis_ElohimFonsecados_M.pdf: 2277988 bytes, checksum: 08f2c3b84a391217d575c0f425159fca (MD5) Previous issue date: 2015
Resumo: Este trabalho é dividido em duas partes. Na primeira parte, uma rede de correlação é construída baseada em um modelo de Ising em diferentes temperaturas, crítica, subcrítica e supercrítica, usando um algorítimo de Metropolis Monte-Carlo com dinâmica de \textit{single-spin-flip}. Este modelo teórico é comparado com uma rede do cérebro construída a partir de correlações das séries temporais do sinal BOLD de fMRI de regiões do cérebro. Medidas de rede, como coeficiente de aglomeração, mínimo caminho médio e distribuição de grau são analisadas. As mesmas medidas de rede são calculadas para a rede obtida pelas correlações das séries temporais dos spins no modelo de Ising. Os resultados da rede cerebral são melhor explicados pelo modelo teórico na temperatura crítica, sugerindo aspectos de criticalidade na dinâmica cerebral. Na segunda parte, é estudada a dinâmica temporal da atividade de um população neural, ou seja, a atividade de células ganglionares da retina gravadas em uma matriz de multi-eletrodos. Vários estudos têm focado em descrever a atividade de redes neurais usando modelos de Ising com desordem, não dando atenção à estrutura dinâmica. Tratando o tempo como uma dimensão extra do sistema, a dinâmica temporal da atividade da população neural é modelada. O princípio de máxima entropia é usado para construir um modelo de Ising com interação entre pares das atividades de diferentes neurônios em tempos diferentes. O ajuste do modelo é feito com uma combinação de amostragem de Monte-Carlo e método do gradiente descendente. O sistema é caracterizado pelos parâmetros aprendidos, questões como balanço detalhado e reversibilidade temporal são analisadas e variáveis termodinâmicas, como o calor específico, podem ser calculadas para estudar aspectos de criticalidade
Abstract: This work is divided in two parts. In the first part, a correlation network is build based on an Ising model at different temperatures, critical, subcritical and supercritical, using a Metropolis Monte-Carlo algorithm with single-spin-flip dynamics. This theoretical model is compared with a brain network built from the correlations of BOLD fMRI temporal series of brain regions activity. Network measures, such as clustering coefficient, average shortest path length and degree distributions are analysed. The same network measures are calculated to the network obtained from the time series correlations of the spins in the Ising model. The results from the brain network are better explained by the theoretical model at the critical temperature, suggesting critical aspects in the brain dynamics. In the second part, the temporal dynamics of the activity of a neuron population, that is, the activity of retinal ganglion cells recorded in a multi-electrode array was studied. Many studies have focused on describing the activity of neural networks using disordered Ising models, with no regard to the dynamic nature. Treating time as an extra dimension of the system, the temporal dynamics of the activity of the neuron population is modeled. The maximum entropy principle approach is used to build an Ising model with pairwise interactions between the activities of different neurons at different times. Model fitting is performed by a combination of Metropolis Monte Carlo sampling with gradient descent methods. The system is characterized by the learned parameters, questions like detailed balance and time reversibility are analysed and thermodynamic variables, such as specific heat, can be calculated to study critical aspects
Mestrado
Física
Mestre em Física
2013/25361-6
FAPESP
31
Varas, Jaime Armando. "Employment of neural networks in the estimation of impact parameters." Thesis, The University of Sydney, 2002. https://hdl.handle.net/2123/27885.
Full textAbstract:
In this study, we employed a two-stage backpropagation neural network (NNW) to estimate
the impact parameter (b) of heavy-ion collisions. Using Monte-Carlo (MC) generated Pb-Pb
events at 160 GeV/nucleon we employed three observables from each event to train the
NNW. The generated events were target-projectile in nature, from which the charged pion
multiplicity (MULT), largest spectator fragment (ZMAX) and charge flow in the forward
direction (sz) were used as input signals for the NNW.
A statistical approach that employed the weighted mean of the three inputs to estimate b was
used as a test method, against which the NNW's results were compared. The results showed,
that the NNW was as accurate as the weighted mean in estimating b.
Using central events fom EMUOl data comprising of Au-Au events at 11 GeV/nucleon and
Pb-Pb events at 158 GeV/nucleon, we extracted the observables MULT, ZMAX and Qm
from each event, and utilising the trained NNW we estimated b.
A comparison of MC generated events of a similar b range was made with our EMUOl
results, and it was shown that the two data sets agreed within statistical errors. A further
comparison of the pseudorapidity distribution between the two data sets revealed that the
estimated b for the EMUOl data, was consistent with the MC data.
32
Battista, Aldo. "Low-dimensional continuous attractors in recurrent neural networks : from statistical physics to computational neuroscience." Thesis, Université Paris sciences et lettres, 2020. http://www.theses.fr/2020UPSLE012.
Full textAbstract:
La manière dont l'information sensorielle est codée et traitée par les circuits neuronaux est une question centrale en neurosciences computationnelles. Dans de nombreuses régions du cerveau, on constate que l'activité des neurones dépend fortement de certains corrélats sensoriels continus ; on peut citer comme exemples les cellules simples de la zone V1 du cortex visuel codant pour l'orientation d'une barre présentée à la rétine, et les cellules de direction de la tête dans le subiculum ou les cellules de lieu dans l'hippocampe, dont les activités dépendent, respectivement, de l'orientation de la tête et de la position d'un animal dans l'espace physique. Au cours des dernières décennies, les réseaux neuronaux à attracteur continu ont été introduits comme un modèle abstrait pour la représentation de quelques variables continues dans une grande population de neurones bruités. Grâce à un ensemble approprié d'interactions par paires entre les neurones, la dynamique du réseau neuronal est contrainte de s'étendre sur une variété de faible dimension dans l'espace de haute dimension des configurations d'activités, et code ainsi quelques coordonnées continues sur la variété, correspondant à des informations spatiales ou sensorielles. Alors que le modèle original était basé sur la construction d'une variété continue unique dans un espace à haute dimension, on s'est vite rendu compte que le même réseau neuronal pouvait coder pour de nombreux attracteurs distincts, correspondant à différents environnements spatiaux ou situations contextuelles. Une solution approximative à ce problème plus difficile a été proposée il y a vingt ans, et reposait sur une prescription ad hoc pour les interactions par paires entre les neurones, résumant les différentes contributions correspondant à chaque attracteur pris indépendamment des autres. Cette solution souffre cependant de deux problèmes majeurs : l'interférence entre les cartes limitent fortement la capacité de stockage, et la résolution spatiale au sein d'une carte n'est pas contrôlée. Dans le présent manuscrit, nous abordons ces deux questions. Nous montrons comment parvenir à un stockage optimal des attracteurs continus et étudions le compromis optimal entre capacité et résolution spatiale, c'est-à-dire comment l'exigence d'une résolution spatiale plus élevée affecte le nombre maximal d'attracteurs pouvant être stockés, prouvant que les réseaux neuronaux récurrents sont des dispositifs de mémoire très efficaces capables de stocker de nombreux attracteurs continus à haute résolution. Afin de résoudre ces problèmes, nous avons utilisé une combinaison de techniques issues de la physique statistique des systèmes désordonnés et de la théorie des matrices aléatoires. D'une part, nous avons étendu la théorie de l'apprentissage de Gardner au cas des modèles présentant de fortes corrélations spatiales. D'autre part, nous avons introduit et étudié les propriétés spectrales d'un nouvel ensemble de matrices aléatoires, c'est-à-dire la superposition additive d'un grand nombre de matrices aléatoires euclidiennes indépendantes dans le régime de haute densité. En outre, cette approche définit un cadre concret pour répondre à de nombreuses questions, en lien étroit avec les expériences en cours, liées notamment à la discussion de l'hypothèse du remapping aléatoire et au codage de l'information spatiale et au développement des circuits cérébraux chez les jeunes animaux. Enfin, nous discutons d'un mécanisme possible pour l'apprentissage des attracteurs continus à partir d'images réellesHow sensory information is encoded and processed by neuronal circuits is a central question in computational neuroscience. In many brain areas, the activity of neurons is found to depend strongly on some continuous sensory correlate; examples include simple cells in the V1 area of the visual cortex coding for the orientation of a bar presented to the retina, and head direction cells in the subiculum or place cells in the hippocampus, whose activities depend, respectively, on the orientation of the head and the position of an animal in the physical space. Over the past decades, continuous attractor neural networks were introduced as an abstract model for the representation of a few continuous variables in a large population of noisy neurons. Through an appropriate set of pairwise interactions between the neurons, the dynamics of the neural network is constrained to span a low-dimensional manifold in the high-dimensional space of activity configurations, and thus codes for a few continuous coordinates on the manifold, corresponding to spatial or sensory information. While the original model was based on how to build a single continuous manifold in an high-dimensional space, it was soon realized that the same neural network should code for many distinct attractors, {em i.e.}, corresponding to different spatial environments or contextual situations. An approximate solution to this harder problem was proposed twenty years ago, and relied on an ad hoc prescription for the pairwise interactions between neurons, summing up the different contributions corresponding to each single attractor taken independently of the others. This solution, however, suffers from two major issues: the interference between maps strongly limit the storage capacity, and the spatial resolution within a map is not controlled. In the present manuscript, we address these two issues. We show how to achieve optimal storage of continuous attractors and study the optimal trade-off between capacity and spatial resolution, that is, how the requirement of higher spatial resolution affects the maximal number of attractors that can be stored, proving that recurrent neural networks are very efficient memory devices capable of storing many continuous attractors at high resolution. In order to tackle these problems we used a combination of techniques from statistical physics of disordered systems and random matrix theory. On the one hand we extended Gardner's theory of learning to the case of patterns with strong spatial correlations. On the other hand we introduced and studied the spectral properties of a new ensemble of random matrices, {em i.e.}, the additive superimposition of an extensive number of independent Euclidean random matrices in the high-density regime. In addition, this approach defines a concrete framework to address many questions, in close connection with ongoing experiments, related in particular to the discussion of the random remapping hypothesis and to the coding of spatial information and the development of brain circuits in young animals. Finally, we discuss a possible mechanism for the learning of continuous attractors from real images
33
Ronchi, Emanuele. "Neural Networks Applications and Electronics Development for Nuclear Fusion Neutron Diagnostics." Doctoral thesis, Uppsala universitet, Institutionen för fysik och astronomi, 2009. http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-108583.
Full textAbstract:
This thesis describes the development of electronic modules for fusion neutron spectroscopy as well as several implementations of artificial neural networks (NN) for neutron diagnostics for the Joint European Torus (JET) experimental reactor in England. The electronics projects include the development of two fast light pulser modules based on Light Emitting Diodes (LEDs) for the calibration and stability monitoring of two neutron spectrometers (MPRu and TOFOR) at JET. The particular electronic implementation of the pulsers allowed for operation of the LEDs in the nanosecond time scale, which is typically not well accessible with simpler circuits. Another electronic project consisted of the the development and implementation at JET of 32 high frequency analog signal amplifiers for MPRu. The circuit board layout adopted and the choice of components permitted to achieve bandwidth above 0.5 GHz and low distortion for a wide range of input signals. The successful and continued use of all electronic modules since 2005 until the present day is an indication of their good performance and reliability. The NN applications include pulse shape discrimination (PSD), deconvolution of experimental data and tomographic reconstruction of neutron emissivity profiles for JET. The first study showed that NN can perform neutron/gamma PSD in liquid scintillators significantly better than other conventional techniques, especially for low deposited energy in the detector. The second study demonstrated that NN can be used for statistically efficient deconvolution of neutron energy spectra, with and without parametric neutron spectroscopic models, especially in the region of low counts in the data. The work on tomography provided a simple but effective parametric model for describing neutron emissivity at JET. This was then successfully implemented with NN for fast and automatic tomographic reconstruction of the JET camera data. The fast execution time of NN, i.e. usually in the microsecond time scale, makes the NN applications presented here suitable for real-time data analysis and typically orders of magnitudes faster than other commonly used codes. The results and numerical methods described in this thesis can be applied to other diagnostic instruments and are of relevance for future fusion reactors such as ITER, currently under construction in Cadarache, France.
34
Sung, Woong Je. "A neural network construction method for surrogate modeling of physics-based analysis." Diss., Georgia Institute of Technology, 2012. http://hdl.handle.net/1853/43721.
Full textAbstract:
A connectivity adjusting learning algorithm, Optimal Brain Growth (OBG) was proposed. Contrast to the conventional training methods for the Artificial Neural Network (ANN) which focus on the weight-only optimization, the OBG method trains both weights and connectivity of a network in a single training process. The standard Back-Propagation (BP) algorithm was extended to exploit the error gradient information of the latent connection whose current weight has zero value. Based on this, the OBG algorithm makes a rational decision between a further adjustment of an existing connection weight and a creation of a new connection having zero weight. The training efficiency of a growing network is maintained by freezing stabilized connections in the further optimization process. A stabilized computational unit is also decomposed into two units and a particular set of decomposition rules guarantees a seamless local re-initialization of a training trajectory. The OBG method was tested for the multiple canonical, regression and classification problems and for a surrogate modeling of the pressure distribution on transonic airfoils. The OBG method showed an improved learning capability in computationally efficient manner compared to the conventional weight-only training using connectivity-fixed Multilayer Perceptrons (MLPs).
35
Pusuluri, Sai Teja. "Exploring Neural Network Models with Hierarchical Memories and Their Use in Modeling Biological Systems." Ohio University / OhioLINK, 2017. http://rave.ohiolink.edu/etdc/view?acc_num=ohiou1490116134938074.
Full text
36
Ericsson, Oscar. "Investigations into neutrino flavor reconstruction from radio detector data using convolutional neural networks." Thesis, Uppsala universitet, Högenergifysik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-449503.
Full textAbstract:
As the IceCube Neutrino Observatory seeks to expand its sensitivity to high PeV-EeV energies by means of the radio technique, the need for fast, efficient and reliable reconstruction methods to recover neutrino properties from radio detector data has emerged. The first recorded investigation into the possibilities of using a neural network based approach to flavor reconstruction is presented. More specifically, a deep convolutional neural network was built and optimized for the purpose of differentiating νe charged current (CC) interaction events from events of all other flavors and interaction modes. The approach is found to be largely successful for neutrino energies above 1018 eV, with a reported accuracy on νe - CC events of > 75% for neutrino energies > 1018.5 eV while maintaining a >60% accuracy for energies > 1018. Predictive accuracy on non- νe - CC events varies between 80% and 90% across the considered neutrino energy range 1017<Eν<1019. The dependence of the accuracy on νe - CC events on neutrino energy is pronounced and attributed to the LPM effect, which alters the features of the radio signals significantly at energies above 1018 eV in contrast to non- νe - CC events. The method shows promise as a first neural network based neutrino flavor reconstruction method, and results can likely be improved through further optimization.
37
Posani, Lorenzo. "Inference and modeling of biological networks : a statistical-physics approach to neural attractors and protein fitness landscapes." Thesis, Paris Sciences et Lettres (ComUE), 2018. http://www.theses.fr/2018PSLEE043/document.
Full textAbstract:
L'avènement récent des procédures expérimentales à haut débit a ouvert une nouvelle ère pour l'étude quantitative des systèmes biologiques. De nos jours, les enregistrements d'électrophysiologie et l'imagerie du calcium permettent l'enregistrement simultané in vivo de centaines à des milliers de neurones. Parallèlement, grâce à des procédures de séquençage automatisées, les bibliothèques de protéines fonctionnelles connues ont été étendues de milliers à des millions en quelques années seulement. L'abondance actuelle de données biologiques ouvre une nouvelle série de défis aux théoriciens. Des méthodes d’analyse précises et transparentes sont nécessaires pour traiter cette quantité massive de données brutes en observables significatifs. Parallèlement, l'observation simultanée d'un grand nombre d'unités en interaction permet de développer et de valider des modèles théoriques visant à la compréhension mécanistique du comportement collectif des systèmes biologiques. Dans ce manuscrit, nous proposons une approche de ces défis basée sur des méthodes et des modèles issus de la physique statistique, en développent et appliquant ces méthodes au problèmes issu de la neuroscience et de la bio-informatique : l’étude de la mémoire spatiale dans le réseau hippocampique, et la reconstruction du paysage adaptatif local d'une protéineThe recent advent of high-throughput experimental procedures has opened a new era for the quantitative study of biological systems. Today, electrophysiology recordings and calcium imaging allow for the in vivo simultaneous recording of hundreds to thousands of neurons. In parallel, thanks to automated sequencing procedures, the libraries of known functional proteins expanded from thousands to millions in just a few years. This current abundance of biological data opens a new series of challenges for theoreticians. Accurate and transparent analysis methods are needed to process this massive amount of raw data into meaningful observables. Concurrently, the simultaneous observation of a large number of interacting units enables the development and validation of theoretical models aimed at the mechanistic understanding of the collective behavior of biological systems. In this manuscript, we propose an approach to both these challenges based on methods and models from statistical physics. We present an application of these methods to problems from neuroscience and bioinformatics, focusing on (1) the spatial memory and navigation task in the hippocampal loop and (2) the reconstruction of the fitness landscape of proteins from homologous sequence data
38
Grose, Mitchell. "Forecasting Atmospheric Turbulence Conditions From Prior Environmental Parameters Using Artificial Neural Networks: An Ensemble Study." University of Dayton / OhioLINK, 2021. http://rave.ohiolink.edu/etdc/view?acc_num=dayton1619632748733788.
Full text
39
Canaday, Daniel M. "Modeling and Control of Dynamical Systems with Reservoir Computing." The Ohio State University, 2019. http://rave.ohiolink.edu/etdc/view?acc_num=osu157469471458874.
Full text
40
Meseguer, Brocal Gabriel. "Multimodal analysis : informed content estimation and audio source separation." Electronic Thesis or Diss., Sorbonne université, 2020. http://www.theses.fr/2020SORUS111.
Full textAbstract:
Cette thèse propose l'étude de l'apprentissage multimodal dans le contexte de signaux musicaux. Tout au long de ce manuscrit, nous nous concentrerons sur l'interaction entre les signaux audio et les informations textuelles. Parmi les nombreuses sources de texte liées à la musique qui peuvent être utilisées (par exemple les critiques, les métadonnées ou les commentaires des réseaux sociaux), nous nous concentrerons sur les paroles. La voix chantée relie directement le signal audio et les informations textuelles d'une manière unique, combinant mélodie et paroles où une dimension linguistique complète l'abstraction des instruments de musique. Notre étude se focalise sur l'interaction audio et paroles pour cibler la séparation de sources et l'estimation de contenu informé. Les stimuli du monde réel sont produits par des phénomènes complexes et leur interaction constante dans divers domaines. Notre compréhension apprend des abstractions utiles qui fusionnent différentes modalités en une représentation conjointe. L'apprentissage multimodal décrit des méthodes qui analysent les phénomènes de différentes modalités et leur interaction afin de s'attaquer à des tâches complexes. Il en résulte des représentations meilleures et plus riches qui améliorent les performances des méthodes d'apprentissage automatique actuelles. Pour développer notre analyse multimodale, nous devons d'abord remédier au manque de données contenant une voix chantée avec des paroles alignées. Ces données sont obligatoires pour développer nos idées. Par conséquent, nous étudierons comment créer une telle base de données en exploitant automatiquement les ressources du World Wide Web. La création de ce type de base de données est un défi en soi qui soulève de nombreuses questions de recherche. Nous travaillons constamment avec le paradoxe classique de la `` poule ou de l'œuf '': l'acquisition et le nettoyage de ces données nécessitent des modèles précis, mais il est difficile de former des modèles sans données. Nous proposons d'utiliser le paradigme enseignant-élève pour développer une méthode où la création de bases de données et l'apprentissage de modèles ne sont pas considérés comme des tâches indépendantes mais plutôt comme des efforts complémentaires. Dans ce processus, les paroles et les annotations non-expertes de karaoké décrivent les paroles comme une séquence de notes alignées sur le temps avec leurs informations textuelles associées. Nous lions ensuite chaque annotation à l'audio correct et alignons globalement les annotations dessusThis dissertation proposes the study of multimodal learning in the context of musical signals. Throughout, we focus on the interaction between audio signals and text information. Among the many text sources related to music that can be used (e.g. reviews, metadata, or social network feedback), we concentrate on lyrics. The singing voice directly connects the audio signal and the text information in a unique way, combining melody and lyrics where a linguistic dimension complements the abstraction of musical instruments. Our study focuses on the audio and lyrics interaction for targeting source separation and informed content estimation. Real-world stimuli are produced by complex phenomena and their constant interaction in various domains. Our understanding learns useful abstractions that fuse different modalities into a joint representation. Multimodal learning describes methods that analyse phenomena from different modalities and their interaction in order to tackle complex tasks. This results in better and richer representations that improve the performance of the current machine learning methods. To develop our multimodal analysis, we need first to address the lack of data containing singing voice with aligned lyrics. This data is mandatory to develop our ideas. Therefore, we investigate how to create such a dataset automatically leveraging resources from the World Wide Web. Creating this type of dataset is a challenge in itself that raises many research questions. We are constantly working with the classic ``chicken or the egg'' problem: acquiring and cleaning this data requires accurate models, but it is difficult to train models without data. We propose to use the teacher-student paradigm to develop a method where dataset creation and model learning are not seen as independent tasks but rather as complementary efforts. In this process, non-expert karaoke time-aligned lyrics and notes describe the lyrics as a sequence of time-aligned notes with their associated textual information. We then link each annotation to the correct audio and globally align the annotations to it. For this purpose, we use the normalized cross-correlation between the voice annotation sequence and the singing voice probability vector automatically, which is obtained using a deep convolutional neural network. Using the collected data we progressively improve that model. Every time we have an improved version, we can in turn correct and enhance the data
41
SAGLIETTI, LUCA. "Out of equilibrium Statistical Physics of learning." Doctoral thesis, Politecnico di Torino, 2018. http://hdl.handle.net/11583/2710532.
Full textAbstract:
In the study of hard optimization problems, it is often unfeasible to achieve
a full analytic control on the dynamics of the algorithmic processes that
find solutions efficiently. In many cases, a static approach is able to provide
considerable insight into the dynamical properties of these algorithms: in fact,
the geometrical structures found in the energetic landscape can strongly affect
the stationary states and the optimal configurations reached by the solvers.
In this context, a classical Statistical Mechanics approach, relying on the
assumption of the asymptotic realization of a Boltzmann Gibbs equilibrium,
can yield misleading predictions when the studied algorithms comprise some
stochastic components that effectively drive these processes out of equilibrium.
Thus, it becomes necessary to develop some intuition on the relevant features
of the studied phenomena and to build an ad hoc Large Deviation analysis,
providing a more targeted and richer description of the geometrical properties
of the landscape. The present thesis focuses on the study of learning processes
in Artificial Neural Networks, with the aim of introducing an out of equilibrium
statistical physics framework, based on the introduction of a local entropy
potential, for supporting and inspiring algorithmic improvements in the field
of Deep Learning, and for developing models of neural computation that can
carry both biological and engineering interest.
42
Marginean, Radu. "Measurement of the top pair production cross section at CDF using neural networks." Connect to this title online, 2004. http://rave.ohiolink.edu/etdc/view?acc%5Fnum=osu1101831484.
Full textAbstract:
Thesis (Ph. D.)--Ohio State University, 2004.Title from first page of PDF file. Document formatted into pages; contains xiii, 110 p.; also includes graphics (some col.). Includes bibliographical references (p. 106-110).
43
Zhao, Ruiguang. "Development of a CMOS pixel sensor with on-chip artificial neural networks." Thesis, Strasbourg, 2019. http://www.theses.fr/2019STRAE050.
Full textAbstract:
Dans le détecteur de vertex de l'ILC (International Linear Collider), un nombre élevé d'impacts supplémentaires seront générés par des électrons résultant de processus liés au bruit de fond des faisceaux. Leur impulsion se trouve typiquement est inférieure à celle des particules issues d'événements associés à des processus physiques. Notre groupe à l'IPHC a proposé d'explorer le concept d'un capteur à pixels CMOS avec des ANNs intégrés pour marquer et supprimer les pixels touchés (hits) générés par ces électrons.Au cours de ma thèse de doctorat, je me suis concentré sur l'étude d'un capteur à pixels CMOS avec des ANNs intégrés portant sur les aspects suivants :1. L'implémentation de modules de prétraitement et d'un ANN dans un composant FPGA pour l'étude de faisabilité ; 2. Un algorithme pour la recherche de clusters, qui fait partie des modules de prétraitement, a été proposé en vue d'être intégré dans la conception de l'ASICIn the vertex detector of the ILC (International Linear Collider), a large number of extra hits will be generated by electrons coming from the beam background. Momenta of these background electrons typically are lower than particles coming from physics events. Our group in IPHC has proposed the concept of a CMOS pixel sensor with on-chip ANNs to tag and remove hits generated by background particles.During my PhD thesis, I focused on the study of a CMOS pixel sensor with on-chip ANNs from the following aspects :1. The implementation of preprocessing modules and an ANN in an FPGA device for the feasibility study ;2. An on-chip algorithm for cluster search which is a part of preprocessing modules has been proposed to integrate into the ASIC design
44
Grassia, Filippo. "Silicon neural networks : implementation of cortical cells to improve the artificial-biological hybrid technique." Phd thesis, Université Sciences et Technologies - Bordeaux I, 2013. http://tel.archives-ouvertes.fr/tel-00789406.
Full textAbstract:
This work has been supported by the European FACETS-ITN project. Within the frameworkof this project, we contribute to the simulation of cortical cell types (employingexperimental electrophysiological data of these cells as references), using a specific VLSIneural circuit to simulate, at the single cell level, the models studied as references in theFACETS project. The real-time intrinsic properties of the neuromorphic circuits, whichprecisely compute neuron conductance-based models, will allow a systematic and detailedexploration of the models, while the physical and analog aspect of the simulations, as opposedthe software simulation aspect, will provide inputs for the development of the neuralhardware at the network level. The second goal of this thesis is to contribute to the designof a mixed hardware-software platform (PAX), specifically designed to simulate spikingneural networks. The tasks performed during this thesis project included: 1) the methodsused to obtain the appropriate parameter sets of the cortical neuron models that can beimplemented in our analog neuromimetic chip (the parameter extraction steps was validatedusing a bifurcation analysis that shows that the simplified HH model implementedin our silicon neuron shares the dynamics of the HH model); 2) the fully customizablefitting method, in voltage-clamp mode, to tune our neuromimetic integrated circuits usinga metaheuristic algorithm; 3) the contribution to the development of the PAX systemin terms of software tools and a VHDL driver interface for neuron configuration in theplatform. Finally, it also addresses the issue of synaptic tuning for future SNN simulation.
45
Colombini, Giulio. "Synchronisation phenomena in complex neuronal networks." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2021. http://amslaurea.unibo.it/23904/.
Full textAbstract:
The phenomenon of neural synchronisation, a simultaneous and repeated firing of clusters of neurons, underlies many physiological functions and pathological manifestations in the brain of humans and animals, ranging from information encoding to epileptic seizures. Neural synchronisation, as a general phenomenon, can be approached theoretically in the framework of Dynamical Systems on Networks. In the present work, we do so by considering complex networks of FitzHugh-Nagumo model neurons. In the first part we consider the most understood models where each neuron treats its presynaptic neurons all on an equal footing, normalising signals with its in-degree. We study the stability of the synchronous state by devising an algorithm that destabilises it by selecting and removing links from the network, so to obtain a bipartite network. The selection is performed using a perturbative expression, which can be regarded as a specialisation of a previously introduced Spectral Centrality measure. The algorithm is tested on Erdős-Renyi, Watts-Strogatz and Barabási-Albert networks, and its behaviour is assessed from a dynamical and from a structural point of view. In the second part we consider the less studied case in which each neuron divides equally its output among the postsynaptic neurons, so to reproduce schematically the situation where a fixed quantity of neurotransmitter is subdivided between several efferent neurons. In this context a self-consistent approach is formulated and its limitations are explored. In order to extend its application to larger networks, a Mean Field Approximation is presented. The predictivity of the Mean Field Approach is then tested on the different random network models, and the results are discussed in terms of the original network properties.
46
Vorvolakos, Angelos. "Artificial neural network methods in high energy physics and their application to the identification of quark and gluon jets in electroproton collisions." Thesis, Imperial College London, 1999. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.314217.
Full text
47
Deans, Christopher Scott. "Closure tested parton distributions for the LHC." Thesis, University of Edinburgh, 2016. http://hdl.handle.net/1842/20375.
Full textAbstract:
Parton distribution functions (PDFs) provide a description of the quark and gluon content of the proton. They are important input into theoretical calculations of hadronic observables, and are obtained by fitting to a wide range of experimental data. The NNPDF approach to fitting PDFs provides a robust and reliable determination of their central values and uncertainties. The PDFs are modelled using neural networks, while the uncertainties are generated through the use of Monte Carlo replica datasets. In this thesis I provide an in depth description of development of the latest NNPDF determination: NNPDF3.0. A number of novel adaptations to the genetic algorithm and network structure are outlined and the results of tests as to their effectiveness are shown. Centrally, the use of closure tests, where artificial data is generated according to a known theory and used to perform a fit, has been instrumental in both the development and validation of the NNPDF3.0 approach. The results of these tests, which demonstrate the ability of our methodology to reproduce a known underlying law, are investigated in detail. Finally, results from the NNPDF3.0 PDF sets are presented. The parton distributions obtained are compared with results from other PDF collaborations, and PDFs fit to limited datasets are also discussed. Physical observables relevant for future collider runs are presented and compared to other determinations.
48
Meenakshisundaram, Venkatesh. "ELUCIDATING PHYSICS OF SEQUENCE-SPECIFIC POLYMERS AND THE GLASS TRANSITION VIA EVOLUTIONARY COMPUTATIONAL DESIGN." University of Akron / OhioLINK, 2018. http://rave.ohiolink.edu/etdc/view?acc_num=akron1513717453745275.
Full text
49
Gullstrand, Mattias, and Stefan Maraš. "Using Graph Neural Networks for Track Classification and Time Determination of Primary Vertices in the ATLAS Experiment." Thesis, KTH, Matematisk statistik, 2020. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-288505.
Full textAbstract:
Starting in 2027, the high-luminosity Large Hadron Collider (HL-LHC) will begin operation and allow higher-precision measurements and searches for new physics processes between elementary particles. One central problem that arises in the ATLAS detector when reconstructing event information is to separate the rare and interesting hard scatter (HS) interactions from uninteresting pileup (PU) interactions in a spatially compact environment. This problem becomes even harder to solve at higher luminosities. This project relies on leveraging the time dimension and determining a time of the HS interactions to separate them from PU interactions by using information measured by the upcoming High-Granularity Timing Detector (HGTD). The current method relies on using a boosted decision tree (BDT) together with the timing information from the HGTD to determine a time. We suggest a novel approach of utilizing a graph attentional network (GAT) where each bunch-crossing is represented as a graph of tracks and the properties of the GAT are applied on a track level to inspect if such a model can outperform the current BDT. Our results show that we are able to replicate the results of the BDT and even improve some metrics at the expense of increasing the uncertainty of the time determination. We conclude that although there is potential for GATs to outperform the BDT, a more complex model should be applied. Finally, we provide some suggestions for improvement and hope to inspire further study and advancements in this direction which shows promising potential.Från och med 2027 kommer \textit{high-luminosity Large Hadron Collider} (HL-LHC) att tas i drift och möjliggöra mätningar med högre precision och utforskningar av nya fysikprocesser mellan elementarpartiklar. Ett centralt problem som uppstår i ATLAS-detektorn vid rekonstruktionen av partikelkollisioner är att separera sällsynta och intressanta interaktioner, så kallade \textit{hard-scatters} (HS) från ointressanta \textit{pileup}-interaktioner (PU) i den kompakta rumsliga dimensionen. Svårighetsgraden för detta problem ökar vid högre luminositeter. Med hjälp av den kommande \textit{High-Granularity Timing-detektorns} (HGTD) mätningar kommer även tidsinformation relaterat till interaktionerna att erhållas. I detta projekt används denna information för att beräkna tiden för enskillda interaktioner vilket därmed kan användas för att separera HS-interaktioner från PU-interaktioner. Den nuvarande metoden använder en trädregressionsmetod, s.k. boosted decision tree (BDT) tillsammans med tidsinformationen från HGTD för att bestämma en tid. Vi föreslår ett nytt tillvägagångssätt baserat på ett s.k. uppvaktande grafnätverk (GAT), där varje protonkollision representeras som en graf över partikelspåren och där GAT-egenskaperna tillämpas på spårnivå. Våra resultat visar att vi kan replikera de BDT-baserade resultaten och till och med förbättra resultaten på bekostnad av att öka osäkerheten i tidsbestämningarna. Vi drar slutsatsen att även om det finns potential för GAT-modeller att överträffa BDT-modeller, bör mer komplexa versioner av de förra tillämpas. Vi ger slutligen några förbättringsförslag som vi hoppas ska kunna inspirera till ytterligare studier och framsteg inom detta område, vilket visar lovande potential.
50
Mukherjee, Rajaditya. "Accelerating Data-driven Simulations for Deformable Bodies and Fluids." The Ohio State University, 2018. http://rave.ohiolink.edu/etdc/view?acc_num=osu1523634514740489.
Full text