Showing 1–10 of 10 results for author: Scholz, F

Quick and Accurate Affordance Learning

Authors: Fedor Scholz, Erik Ayari, Johannes Bertram, Martin V. Butz

Abstract: Infants learn actively in their environments, sha** their own learning curricula. They learn about their environments' affordances, that is, how local circumstances determine how their behavior can affect the environment. Here we model this type of behavior by means of a deep learning architecture. The architecture mediates between global cognitive map exploration and local affordance learning.… ▽ More Infants learn actively in their environments, sha** their own learning curricula. They learn about their environments' affordances, that is, how local circumstances determine how their behavior can affect the environment. Here we model this type of behavior by means of a deep learning architecture. The architecture mediates between global cognitive map exploration and local affordance learning. Inference processes actively move the simulated agent towards regions where they expect affordance-related knowledge gain. We contrast three measures of uncertainty to guide this exploration: predicted uncertainty of a model, standard deviation between the means of several models (SD), and the Jensen-Shannon Divergence (JSD) between several models. We show that the first measure gets fooled by aleatoric uncertainty inherent in the environment, while the two other measures focus learning on epistemic uncertainty. JSD exhibits the most balanced exploration strategy. From a computational perspective, our model suggests three key ingredients for coordinating the active generation of learning curricula: (1) Navigation behavior needs to be coordinated with local motor behavior for enabling active affordance learning. (2) Affordances need to be encoded locally for acquiring generalized knowledge. (3) Effective active affordance learning mechanisms should use density comparison techniques for estimating expected knowledge gain. Future work may seek collaborations with developmental psychology to model active play in children in more realistic scenarios. △ Less

Submitted 13 May, 2024; originally announced May 2024.

Adaptive optimization of isogeometric multi-patch discretizations using artificial neural networks

Authors: Dany Rios, Felix Scholz, Thomas Takacs

Abstract: In isogeometric analysis, isogeometric function spaces are employed for accurately representing the solution to a partial differential equation (PDE) on a parameterized domain. They are generated from a tensor-product spline space by composing the basis functions with the inverse of the parameterization. Depending on the geometry of the domain and on the data of the PDE, the solution might not hav… ▽ More In isogeometric analysis, isogeometric function spaces are employed for accurately representing the solution to a partial differential equation (PDE) on a parameterized domain. They are generated from a tensor-product spline space by composing the basis functions with the inverse of the parameterization. Depending on the geometry of the domain and on the data of the PDE, the solution might not have maximum Sobolev regularity, leading to a reduced convergence rate. In this case it is necessary to reduce the local mesh size close to the singularities. The classical approach is to perform adaptive h-refinement, which either leads to an unnecessarily large number of degrees of freedom or to a spline space that does not possess a tensor-product structure. Based on the concept of r-adaptivity we present a novel approach for finding a suitable isogeometric function space for a given PDE without sacrificing the tensor-product structure of the underlying spline space. In particular, we use the fact that different reparameterizations of the same computational domain lead to different isogeometric function spaces while preserving the geometry. Starting from a multi-patch domain consisting of bilinearly parameterized patches, we aim to find the biquadratic multi-patch parameterization that leads to the isogeometric function space with the smallest best approximation error of the solution. In order to estimate the location of the optimal control points, we employ a trained residual neural network that is applied to the graph surfaces of the approximated solution and its derivatives. In our experimental results, we observe that our new method results in a vast improvement of the approximation error for different PDE problems on multi-patch domains. △ Less

Submitted 28 March, 2024; originally announced March 2024.

MSC Class: 65N50

Multi-Mode Front Lens for Momentum Microscopy: Part II Experiments

Authors: O. Tkach, S. Fragkos, Q. Nguyen, S. Chernov, M. Scholz, N. Wind, S. Babenkov, O. Fedchenko, Y. Lytvynenko, D. Zimmer, A. Hloskovskii, D. Kutnyakhov, F. Pressacco, J. Dilling, L. Bruckmeier, M. Heber, F. Scholz, J. Sobota, J. Koralek, N. Sirica, M. Kallmayer, M. Hoesch, C. Schlueter, L. V. Odnodvorets, Y. Mairesse , et al. (4 additional authors not shown)

Abstract: We have experimentally demonstrated different operating modes for the front lenses of the momentum microscopes described in Part I. Measurements at energies from vacuum UV at a high-harmonic generation (HHG)-based source to the soft and hard X-ray range at a synchrotron facility validated the results of theoretical ray-tracing calculations. The key element is a ring electrode concentric with the e… ▽ More We have experimentally demonstrated different operating modes for the front lenses of the momentum microscopes described in Part I. Measurements at energies from vacuum UV at a high-harmonic generation (HHG)-based source to the soft and hard X-ray range at a synchrotron facility validated the results of theoretical ray-tracing calculations. The key element is a ring electrode concentric with the extractor electrode, which can tailor the field in the gap. First, the gap-lens-assisted extractor mode reduces the field strength at the sample while mitigating image aberrations. This mode gave good results in all spectral ranges. Secondly, by compensating the field at the sample surface with a negative voltage at the ring electrode we can operate in zero-field mode, which is beneficial for operando experiments. Finally, higher negative voltages establish the repeller mode, which removes all slow electrons below a certain kinetic energy to eliminate the primary contribution to the space-charge interaction in pump-probe experiments. The switch from extractor to repeller mode is associated with a reduction in the k-field-of-view (10-20 % at hard-X-ray energies, increasing to ~50% at low energies). Real-space imaging also benefits from the new lens modes as confirmed by ToF-XPEEM imaging with 650 nm resolution. △ Less

Submitted 19 January, 2024; v1 submitted 18 January, 2024; originally announced January 2024.

Comments: 22 pages, 9 figures, 56 references

On Steel's Test with Ties

Authors: F. W. Scholz

Abstract: This note revisits Steel's multiple comparison test which uses Wilcoxon statistics in pairwise comparisons of several treatment samples with a common control sample. It derives means, variances and covariances of the Wilcoxon statistics under the conditional randomization distribution, given the tie pattern in the pooled samples. Sample sizes do not have to be equal. Under the randomization distri… ▽ More This note revisits Steel's multiple comparison test which uses Wilcoxon statistics in pairwise comparisons of several treatment samples with a common control sample. It derives means, variances and covariances of the Wilcoxon statistics under the conditional randomization distribution, given the tie pattern in the pooled samples. Sample sizes do not have to be equal. Under the randomization distribution asymptotic multivariate normality of these Wilcoxon statistics is established. This widens the scope of normal approximations to conditional tests, assuming independent samples of respective sizes $n_0, n_1,\ldots,n_K$ from any common population or randomized treatment assignment to $N=n_0 + n_1+\ldots+n_K$ experimental subjects. Significance probabilities are obtained using a normal approximation and a single quadrature. In the continuous shift model the simultaneous tests are converted to simultaneous confidence bounds and intervals. This is implemented in the {\sf R} package {\tt kSamples}. Extensions to all pairwise Wilcoxon test comparisons are discussed. △ Less

Submitted 10 August, 2023; originally announced August 2023.

Locally refined quad meshing for linear elasticity problems based on convolutional neural networks

Authors: Chiu Ling Chan, Felix Scholz, Thomas Takacs

Abstract: In this paper we propose a method to generate suitably refined finite element meshes using neural networks. As a model problem we consider a linear elasticity problem on a planar domain (possibly with holes) having a polygonal boundary. We impose boundary conditions by fixing the position of a part of the boundary and applying a force on another part of the boundary. The resulting displacement and… ▽ More In this paper we propose a method to generate suitably refined finite element meshes using neural networks. As a model problem we consider a linear elasticity problem on a planar domain (possibly with holes) having a polygonal boundary. We impose boundary conditions by fixing the position of a part of the boundary and applying a force on another part of the boundary. The resulting displacement and distribution of stresses depend on the geometry of the domain and on the boundary conditions. When applying a standard Galerkin discretization using quadrilateral finite elements, one usually has to perform adaptive refinement to properly resolve maxima of the stress distribution. Such an adaptive scheme requires a local error estimator and a corresponding local refinement strategy. The overall costs of such a strategy are high. We propose to reduce the costs of obtaining a suitable discretization by training a neural network whose evaluation replaces this adaptive refinement procedure. We set up a single network for a large class of possible domains and boundary conditions and not on a single domain of interest. The computational domain and boundary conditions are interpreted as images, which are suitable inputs for convolution neural networks. We use the U-net architecture and we devise training strategies by dividing the possible inputs into different categories based on their overall geometric complexity. Thus, we compare different training strategies based on varying geometric complexity. One of the advantages of the proposed approach is the interpretation of input and output as images, which do not depend on the underlying discretization scheme. Another is the generalizability and geometric flexibility. The network can be applied to previously unseen geometries, even with different topology and level of detail. Thus, training can easily be extended to other classes of geometries. △ Less

Submitted 15 March, 2022; originally announced March 2022.

MSC Class: 65N50

Inference of Affordances and Active Motor Control in Simulated Agents

Authors: Fedor Scholz, Christian Gumbsch, Sebastian Otte, Martin V. Butz

Abstract: Flexible, goal-directed behavior is a fundamental aspect of human life. Based on the free energy minimization principle, the theory of active inference formalizes the generation of such behavior from a computational neuroscience perspective. Based on the theory, we introduce an output-probabilistic, temporally predictive, modular artificial neural network architecture, which processes sensorimotor… ▽ More Flexible, goal-directed behavior is a fundamental aspect of human life. Based on the free energy minimization principle, the theory of active inference formalizes the generation of such behavior from a computational neuroscience perspective. Based on the theory, we introduce an output-probabilistic, temporally predictive, modular artificial neural network architecture, which processes sensorimotor information, infers behavior-relevant aspects of its world, and invokes highly flexible, goal-directed behavior. We show that our architecture, which is trained end-to-end to minimize an approximation of free energy, develops latent states that can be interpreted as affordance maps. That is, the emerging latent states signal which actions lead to which effects dependent on the local context. In combination with active inference, we show that flexible, goal-directed behavior can be invoked, incorporating the emerging affordance maps. As a result, our simulated agent flexibly steers through continuous spaces, avoids collisions with obstacles, and prefers pathways that lead to the goal with high certainty. Additionally, we show that the learned agent is highly suitable for zero-shot generalization across environments: After training the agent in a handful of fixed environments with obstacles and other terrains affecting its behavior, it performs similarly well in procedurally generated environments containing different amounts of obstacles and terrains of various sizes at different locations. △ Less

Submitted 2 August, 2022; v1 submitted 23 February, 2022; originally announced February 2022.

Comments: 26 pages, 12 figures, submitted to Frontiers in Neurorobotics

Recovery of high-energy photoelectron circular dichroism through Fano interference

Authors: G. Hartmann, M. Ilchen, Ph. Schmidt, C. Küstner-Wetekam, C. Ozga, F. Scholz, J. Buck, F. Trinter, J. Viefhaus, A. Ehresmann, M. S. Schöffler, A. Knie, Ph. V. Demekhin

Abstract: It is commonly accepted that the magnitude of a photoelectron circular dichroism (PECD) is governed by the ability of an outgoing photoelectron wave packet to probe the chiral asymmetry of a molecule. To be able to accumulate this characteristic asymmetry while esca** the chiral ion, photoelectrons need to have relatively small kinetic energies of up to a few tens of electron volts. Here, we dem… ▽ More It is commonly accepted that the magnitude of a photoelectron circular dichroism (PECD) is governed by the ability of an outgoing photoelectron wave packet to probe the chiral asymmetry of a molecule. To be able to accumulate this characteristic asymmetry while esca** the chiral ion, photoelectrons need to have relatively small kinetic energies of up to a few tens of electron volts. Here, we demonstrate a substantial PECD for very fast photoelectrons above 500 eV kinetic energy released from methyloxirane by a participator resonant Auger decay of its lowermost O $1s$-excitation. This effect emerges as a result of the Fano interference between the direct and resonant photoionization pathways, notwithstanding that their individual effects are negligibly small. The resulting dichroic parameter has an anomalous dispersion, i.e. it changes its sign across the resonance, which can be considered as an analogue of the Cotton effect in the X-ray regime. △ Less

Submitted 25 June, 2019; originally announced June 2019.

Comments: 2 figures

Journal ref: Phys. Rev. Lett. 123, 043202 (2019)

Acoustic properties of glacial ice for neutrino detection and the Enceladus Explorer

Authors: K. Helbing, R. Hoffmann, U. Naumann, D. Eliseev, D. Heinen, F. Scholz, C. Wiebusch, S. Zierke

Abstract: Ultra high energy neutrinos may be observed in ice by the emission of acoustic signals. The SPATS detector has investigated the possibility of observing GZK-neutrinos in the clear ice near the South Pole at the IceCube detector site. To explore other potential detection sites glacial ice in the Alps and in Antarctica has been surveyed for its acoustical properties. The purpose of the Enceladus Exp… ▽ More Ultra high energy neutrinos may be observed in ice by the emission of acoustic signals. The SPATS detector has investigated the possibility of observing GZK-neutrinos in the clear ice near the South Pole at the IceCube detector site. To explore other potential detection sites glacial ice in the Alps and in Antarctica has been surveyed for its acoustical properties. The purpose of the Enceladus Explorer (EnEx), on the other hand, is the search for extraterrestrial life on the Saturn moon Enceladus. Here acoustics is used to maneuver a subsurface probe inside the ice by trilateration of signals. A system of acoustic transducers has been developed to study both applications. In the south polar region of the moon Enceladus there are secluded crevasses. These are filled with liquid water, probably heated by tidal forces due to the short distance to Saturn. We intend to take a sample of water from these crevasses by using a combination of a melt down and steering probe called IceMole (IM). Maneuvering IM requires a good understanding of ice properties such as the speed of sound, the attenuation of acoustic signals in ice, their directional dependencies and their dependence on different frequencies. The technology developed for this positioning system could also contribute to the design of future large scale acoustic neutrino detectors. We present our analysis methods and the findings on attenuation, sound speed, and frequency response obtained at several sites in the Alps and Antarctica. △ Less

Submitted 17 August, 2016; originally announced August 2016.

Comments: 7 pages, 12 figures, ICRC 2015

Imaging Molecular Structure through Femtosecond Photoelectron Diffraction on Aligned and Oriented Gas-Phase Molecules

Authors: R. Boll, A. Rouzee, M. Adolph, D. Anielski, A. Aquila, S. Bari, C. Bomme, C. Bostedt, J. D. Bozek, H. N. Chapman, L. Christensen, R. Coffee, N. Coppola, S. De, P. Decleva, S. W. Epp, B. Erk, F. Filsinger, L. Foucar, T. Gorkhover, L. Gumprecht, A. Hoemke, L. Holmegaard, P. Johnsson, J. S. Kienitz , et al. (27 additional authors not shown)

Abstract: This paper gives an account of our progress towards performing femtosecond time-resolved photoelectron diffraction on gas-phase molecules in a pump-probe setup combining optical lasers and an X-ray Free-Electron Laser. We present results of two experiments aimed at measuring photoelectron angular distributions of laser-aligned 1-ethynyl-4-fluorobenzene (C8H5F) and dissociating, laseraligned 1,4-di… ▽ More This paper gives an account of our progress towards performing femtosecond time-resolved photoelectron diffraction on gas-phase molecules in a pump-probe setup combining optical lasers and an X-ray Free-Electron Laser. We present results of two experiments aimed at measuring photoelectron angular distributions of laser-aligned 1-ethynyl-4-fluorobenzene (C8H5F) and dissociating, laseraligned 1,4-dibromobenzene (C6H4Br2) molecules and discuss them in the larger context of photoelectron diffraction on gas-phase molecules. We also show how the strong nanosecond laser pulse used for adiabatically laser-aligning the molecules influences the measured electron and ion spectra and angular distributions, and discuss how this may affect the outcome of future time-resolved photoelectron diffraction experiments. △ Less

Submitted 29 July, 2014; originally announced July 2014.

Comments: 24 pages, 10 figures, Faraday Discussions 171

Absolute cross sections for photoionization of Xe$^{q+}$ ions (1 $\le$ q $\le$ 5) at the 3d ionization threshold

Authors: S. Schippers, S. Ricz, T. Buhr, A. Borovik Jr., J. Hellhund, K. Holste, K. Huber, H. -J. Schäfer, D. Schury, S. Klumpp, K. Mertens, M. Martins, R. Flesch, G. Ulrich, E. Rühl, T. Jahnke, J. Lower, D. Metz, L. P. H. Schmidt, M. Schöffler, J. B. Williams, L. Glaser, F. Scholz, J. Seltmann, J. Viefhaus , et al. (4 additional authors not shown)

Abstract: The photon-ion merged-beams technique has been employed at the new Photon-Ion spectrometer at PETRA III (PIPE) for measuring multiple photoionization of Xe$^{q+}$ (q=1-5) ions. Total ionization cross sections have been obtained on an absolute scale for the dominant ionization reactions of the type hν+ Xe$^{q+}$ $\to$ Xe$^{r+}$ + (q-r) e$^-$ with product charge states q+2 $\le$ r $\le$ q+5. Promine… ▽ More The photon-ion merged-beams technique has been employed at the new Photon-Ion spectrometer at PETRA III (PIPE) for measuring multiple photoionization of Xe$^{q+}$ (q=1-5) ions. Total ionization cross sections have been obtained on an absolute scale for the dominant ionization reactions of the type hν+ Xe$^{q+}$ $\to$ Xe$^{r+}$ + (q-r) e$^-$ with product charge states q+2 $\le$ r $\le$ q+5. Prominent ionization features are observed in the photon-energy range 650-750 eV, which are associated with excitation or ionization of an inner-shell 3d electron. Single-configuration Dirac-Fock calculations agree quantitatively with the experimental cross sections for non-resonant photoabsorption, but fail to reproduce all details of the measured ionization resonance structures. △ Less

Submitted 11 April, 2014; originally announced April 2014.

Comments: 20 pages, 12 figures, submitted for publication to J. Phys. B

Journal ref: J. Phys. B. 47 (2014) 115602