004 Datenverarbeitung; Informatik
Refine
Year of publication
Document Type
- Article (19)
- Preprint (7)
- Conference Proceeding (3)
- Doctoral Thesis (3)
- Part of a Book (2)
- diplomthesis (1)
Has Fulltext
- yes (35)
Is part of the Bibliography
- no (35)
Keywords
- synaptic plasticity (4)
- Fisher information (3)
- Hebbian learning (3)
- objective functions (3)
- deep learning (2)
- generating functionals (2)
- homeostatic adaption (2)
- Artificial Intelligence (1)
- CBM detector (1)
- CLVisc (1)
Institute
- Physik (35) (remove)
We introduce a novel technique that utilizes a physics-driven deep learning method to reconstruct the dense matter equation of state from neutron star observables, particularly the masses and radii. The proposed framework involves two neural networks: one to optimize the EoS using Automatic Differentiation in the unsupervised learning scheme; and a pre-trained network to solve the Tolman–Oppenheimer–Volkoff (TOV) equations. The gradient-based optimization process incorporates a Bayesian picture into the proposed framework. The reconstructed EoS is proven to be consistent with the results from conventional methods. Furthermore, the resulting tidal deformation is in agreement with the limits obtained from the gravitational wave event, GW170817.
Recurrent cortical network dynamics plays a crucial role for sequential information processing in the brain. While the theoretical framework of reservoir computing provides a conceptual basis for the understanding of recurrent neural computation, it often requires manual adjustments of global network parameters, in particular of the spectral radius of the recurrent synaptic weight matrix. Being a mathematical and relatively complex quantity, the spectral radius is not readily accessible to biological neural networks, which generally adhere to the principle that information about the network state should either be encoded in local intrinsic dynamical quantities (e.g. membrane potentials), or transmitted via synaptic connectivity. We present two synaptic scaling rules for echo state networks that solely rely on locally accessible variables. Both rules work online, in the presence of a continuous stream of input signals. The first rule, termed flow control, is based on a local comparison between the mean squared recurrent membrane potential and the mean squared activity of the neuron itself. It is derived from a global scaling condition on the dynamic flow of neural activities and requires the separability of external and recurrent input currents. We gained further insight into the adaptation dynamics of flow control by using a mean field approximation on the variances of neural activities that allowed us to describe the interplay between network activity and adaptation as a two-dimensional dynamical system. The second rule that we considered, variance control, directly regulates the variance of neural activities by locally scaling the recurrent synaptic weights. The target set point of this homeostatic mechanism is dynamically determined as a function of the variance of the locally measured external input. This functional relation was derived from the same mean-field approach that was used to describe the approximate dynamics of flow control.
The effectiveness of the presented mechanisms was tested numerically using different external input protocols. The network performance after adaptation was evaluated by training the network to perform a time delayed XOR operation on binary sequences. As our main result, we found that flow control can reliably regulate the spectral radius under different input statistics, but precise tuning is negatively affected by interneural correlations. Furthermore, flow control showed a consistent task performance over a wide range of input strengths/variances. Variance control, on the other side, did not yield the desired spectral radii with the same precision. Moreover, task performance was less consistent across different input strengths.
Given the better performance and simpler mathematical form of flow control, we concluded that a local control of the spectral radius via an implicit adaptation scheme is a realistic alternative to approaches using classical “set point” homeostatic feedback controls of neural firing.
Author summary How can a neural network control its recurrent synaptic strengths such that network dynamics are optimal for sequential information processing? An important quantity in this respect, the spectral radius of the recurrent synaptic weight matrix, is a non-local quantity. Therefore, a direct calculation of the spectral radius is not feasible for biological networks. However, we show that there exist a local and biologically plausible adaptation mechanism, flow control, which allows to control the recurrent weight spectral radius while the network is operating under the influence of external inputs. Flow control is based on a theorem of random matrix theory, which is applicable if inter-synaptic correlations are weak. We apply the new adaption rule to echo-state networks having the task to perform a time-delayed XOR operation on random binary input sequences. We find that flow-controlled networks can adapt to a wide range of input strengths while retaining essentially constant task performance.
Recurrent cortical network dynamics plays a crucial role for sequential information processing in the brain. While the theoretical framework of reservoir computing provides a conceptual basis for the understanding of recurrent neural computation, it often requires manual adjustments of global network parameters, in particular of the spectral radius of the recurrent synaptic weight matrix. Being a mathematical and relatively complex quantity, the spectral radius is not readily accessible to biological neural networks, which generally adhere to the principle that information about the network state should either be encoded in local intrinsic dynamical quantities (e.g. membrane potentials), or transmitted via synaptic connectivity. We present two synaptic scaling rules for echo state networks that solely rely on locally accessible variables. Both rules work online, in the presence of a continuous stream of input signals. The first rule, termed flow control, is based on a local comparison between the mean squared recurrent membrane potential and the mean squared activity of the neuron itself. It is derived from a global scaling condition on the dynamic flow of neural activities and requires the separability of external and recurrent input currents. We gained further insight into the adaptation dynamics of flow control by using a mean field approximation on the variances of neural activities that allowed us to describe the interplay between network activity and adaptation as a two-dimensional dynamical system. The second rule that we considered, variance control, directly regulates the variance of neural activities by locally scaling the recurrent synaptic weights. The target set point of this homeostatic mechanism is dynamically determined as a function of the variance of the locally measured external input. This functional relation was derived from the same mean-field approach that was used to describe the approximate dynamics of flow control.
The effectiveness of the presented mechanisms was tested numerically using different external input protocols. The network performance after adaptation was evaluated by training the network to perform a time delayed XOR operation on binary sequences. As our main result, we found that flow control can reliably regulate the spectral radius under different input statistics, but precise tuning is negatively affected by interneural correlations. Furthermore, flow control showed a consistent task performance over a wide range of input strengths/variances. Variance control, on the other side, did not yield the desired spectral radii with the same precision. Moreover, task performance was less consistent across different input strengths.
Given the better performance and simpler mathematical form of flow control, we concluded that a local control of the spectral radius via an implicit adaptation scheme is a realistic alternative to approaches using classical “set point” homeostatic feedback controls of neural firing.
Author summary How can a neural network control its recurrent synaptic strengths such that network dynamics are optimal for sequential information processing? An important quantity in this respect, the spectral radius of the recurrent synaptic weight matrix, is a non-local quantity. Therefore, a direct calculation of the spectral radius is not feasible for biological networks. However, we show that there exist a local and biologically plausible adaptation mechanism, flow control, which allows to control the recurrent weight spectral radius while the network is operating under the influence of external inputs. Flow control is based on a theorem of random matrix theory, which is applicable if inter-synaptic correlations are weak. We apply the new adaption rule to echo-state networks having the task to perform a time-delayed XOR operation on random binary input sequences. We find that flow-controlled networks can adapt to a wide range of input strengths while retaining essentially constant task performance.
Artificial intelligence in heavy-ion collisions : bridging the gap between theory and experiments
(2023)
Artificial Intelligence (AI) methods are employed to study heavy-ion collisions at intermediate collision energies, where high baryon density and moderate temperature QCD matter is produced. The experimental measurements of various conventional observables such as collective flow, particle number fluctuations, etc. are usually compared with expensive model calculations to infer the physics governing the evolution of the matter produced in the collisions. Various experimental effects and processing algorithms can greatly affect the sensitivity of these observables. AI methods are used to bridge this gap between theory and experiments of heavy-ion collisions. The problems with conventional methods of analyzing experimental data are illustrated in a comparative study of the Glauber MC model and the UrQMD transport model. It is found that the centrality determination and the estimated fluctuations of the number of participant nucleons suffer from strong model dependencies for Au-Au collisions at 1.23 AGeV. This can bias the results of the experimental analysis if the number of participant nucleons used is not consistent throughout the analysis and in the final model-to-data comparison. The measurable consequences of this model dependence of the number of participant nucleons are also discussed. In this context, PointNet-based AI models are developed to accurately reconstruct the impact parameter or the number of participant nucleons in a collision event from the hits and/or reconstructed track of particles in 10 AGeV Au-Au collisions at the CBM experiment. In the last part of the thesis, different AI methods to study the equation of state (EoS) at high baryon densities are discussed. First, a Bayesian inference is performed to constrain the density dependence of the EoS from the available experimental measurements of elliptical flow and mean transverse kinetic energy of mid rapidity protons in intermediate energy collisions. The UrQMD model was augmented to include arbitrary potentials (or equivalently the EoSs) in the QMD part to provide a consistent treatment of the EoS throughout the evolution of the system. The experimental data constrain the posterior constructed for the EoS for densities up to four times saturation density. However, beyond three times saturation density, the shape of the posterior depends on the choice of observables used. There is a tension in the measurements at a collision energy of about 4 GeV. This could indicate large uncertainties in the measurements, or alternatively the inability of the underlying model to describe the observables with a given input EoS. Tighter constraints and fully conclusive statements on the EoS require accurate, high statistics data in the whole beam energy range of 2-10 GeV, which will hopefully be provided by the beam energy scan programme of STAR-FXT at RHIC, the upcoming CBM experiment at FAIR, and future experiments at HIAF and NICA. Finally, it is shown that the PointNet-based models can also be used to identify the equation of state in the CBM experiment. Despite the uncertainties due to limited detector acceptance and biases in the reconstruction algorithms, the PointNet-based models are able to learn the features that can accurately identify the underlying physics of the collision. The PointNet-based models are an ideal AI tool to study heavy-ion collisions, not only to identify the geometric event features, such as the impact parameter or the number of participant nucleons, but also to extract abstract physical features, such as the EoS, directly from the detector outputs.
A deep convolutional neural network (CNN) is developed to study symmetry energy (Esym(ρ)) effects by learning the mapping between the symmetry energy and the two-dimensional (transverse momentum and rapidity) distributions of protons and neutrons in heavy-ion collisions. Supervised training is performed with labeled data-set from the ultrarelativistic quantum molecular dynamics (UrQMD) model simulation. It is found that, by using proton spectra on event-by-event basis as input, the accuracy for classifying the soft and stiff Esym(ρ) is about 60% due to large event-by-event fluctuations, while by setting event-summed proton spectra as input, the classification accuracy increases to 98%. The accuracies for 5-label (5 different Esym(ρ)) classification task are about 58% and 72% by using proton and neutron spectra, respectively. For the regression task, the mean absolute errors (MAE) which measure the average magnitude of the absolute differences between the predicted and actual L (the slope parameter of Esym(ρ)) are about 20.4 and 14.8 MeV by using proton and neutron spectra, respectively. Fingerprints of the density-dependent nuclear symmetry energy on the transverse momentum and rapidity distributions of protons and neutrons can be identified by convolutional neural network algorithm.
The state-of-the-art pattern recognition method in machine learning (deep convolution neural network) is used to identify the equation of state (EoS) employed in the relativistic hydrodynamic simulations of heavy ion collisions. High-level correlations of particle spectra in transverse momentum and azimuthal angle learned by the network act as an effective EoS-meter in deciphering the nature of the phase transition in QCD. The EoS-meter is model independent and insensitive to other simulation inputs including the initial conditions and shear viscosity for hydrodynamic simulations. Through this study we demonstrate that there is a traceable encoder of the dynamical information from the phase structure that survives the evolution and exists in the final snapshot of heavy ion collisions and one can exclusively and effectively decode these information from the highly complex final output with machine learning when traditional methods fail. Besides the deep neural network, the performance of traditional machine learning classifiers are also provided.
In this proceeding, we review our recent work using deep convolutional neural network (CNN) to identify the nature of the QCD transition in a hybrid modeling of heavy-ion collisions. Within this hybrid model, a viscous hydrodynamic model is coupled with a hadronic cascade “after-burner”. As a binary classification setup, we employ two different types of equations of state (EoS) of the hot medium in the hydrodynamic evolution. The resulting final-state pion spectra in the transverse momentum and azimuthal angle plane are fed to the neural network as the input data in order to distinguish different EoS. To probe the effects of the fluctuations in the event-by-event spectra, we explore different scenarios for the input data and make a comparison in a systematic way. We observe a clear hierarchy in the predictive power when the network is fed with the event-by-event, cascade-coarse-grained and event-fine-averaged spectra. The carefully-trained neural network can extract high-level features from pion spectra to identify the nature of the QCD transition in a realistic simulation scenario.
Recent experimental findings have reported the presence of unconventional charge orders in the enlarged (2 × 2) unit-cell of kagome metals AV3Sb5 (A = K, Rb, Cs) and hinted towards specific topological signatures. Motivated by these discoveries, we investigate the types of topological phases that can be realized in such kagome superlattices. In this context, we employ a recently introduced statistical method capable of constructing topological models for any generic lattice. By analyzing large data sets generated from symmetry-guided distributions of randomized tight-binding parameters, and labeled with the corresponding topological index, we extract physically meaningful information. We illustrate the possible real-space manifestations of charge and bond modulations and associated flux patterns for different topological classes, and discuss their relation to present theoretical predictions and experimental signatures for the AV3Sb5 family. Simultaneously, we predict higher-order topological phases that may be realized by appropriately manipulating the currently known systems.
TriMem: A parallelized hybrid Monte Carlo software for efficient simulations of lipid membranes
(2022)
Lipid membranes are integral building blocks of living cells and perform a multitude of biological functions. Currently, molecular simulations of cellular-scale membrane remodeling processes at atomic resolution are extremely difficult, due to their size, complexity, and the large times-scales on which these processes occur. Instead, elastic membrane models are used to simulate membrane shapes and transitions between them and to infer their properties and functions. Unfortunately, an efficiently parallelized open-source simulation code to do so has been lacking. Here, we present TriMem, a parallel hybrid Monte Carlo simulation engine for triangulated lipid membranes. The kernels are efficiently coded in C++ and wrapped with Python for ease-of-use. The parallel implementation of the energy and gradient calculations and of Monte Carlo flip moves of edges in the triangulated membrane enable us to simulate large and highly curved membrane structures. For validation, we reproduce phase diagrams of vesicles with varying surface-to-volume ratios and area difference. We also compute the density of states to verify correct Boltzmann sampling. The software can be used to tackle a range of large-scale membrane remodeling processes as a step toward cell-scale simulations. Additionally, extensive documentation make the software accessible to the broad biophysics and computational cell biology communities.
Tracking influenza a virus infection in the lung from hematological data with machine learning
(2022)
The tracking of pathogen burden and host responses with minimal-invasive methods during respiratory infections is central for monitoring disease development and guiding treatment decisions. Utilizing a standardized murine model of respiratory Influenza A virus (IAV) infection, we developed and tested different supervised machine learning models to predict viral burden and immune response markers, i.e. cytokines and leukocytes in the lung, from hematological data. We performed independently in vivo infection experiments to acquire extensive data for training and testing purposes of the models. We show here that lung viral load, neutrophil counts, cytokines like IFN-γ and IL-6, and other lung infection markers can be predicted from hematological data. Furthermore, feature analysis of the models shows that blood granulocytes and platelets play a crucial role in prediction and are highly involved in the immune response against IAV. The proposed in silico tools pave the path towards improved tracking and monitoring of influenza infections and possibly other respiratory infections based on minimal-invasively obtained hematological parameters.