Parameterizing echo state networks for multi-step time series prediction. - In: Neurocomputing, ISSN 1872-8286, Bd. 522 (2023), S. 214-228
Prediction of multi-dimensional time-series data, which may represent such diverse phenomena as climate changes or financial markets, remains a challenging task in view of inherent nonlinearities and non-periodic behavior In contrast to other recurrent neural networks, echo state networks (ESNs) are attractive for (online) learning due to lower requirements w.r.t.training data and computational power. However, the randomly-generated reservoir renders the choice of suitable hyper-parameters as an open research topic. We systematically derive and exemplarily demonstrate design guidelines for the hyper-parameter optimization of ESNs. For the evaluation, we focus on the prediction of chaotic time series, an especially challenging problem in machine learning. Our findings demonstrate the power of a hyper-parameter-tuned ESN when auto-regressively predicting time series over several hundred steps. We found that ESNs’ performance improved by 85.1%-99.8% over an already wisely chosen default parameter initialization. In addition, the fluctuation range is considerably reduced such that significantly worse performance becomes very unlikely across random reservoir seeds. Moreover, we report individual findings per hyper-parameter partly contradicting common knowledge to further, help researchers when training new models.
https://doi.org/10.1016/j.neucom.2022.11.044
Precise motor mapping with transcranial magnetic stimulation. - In: Nature protocols, ISSN 1750-2799, Bd. 18 (2023), S. 293-318
We describe a routine to precisely localize cortical muscle representations within the primary motor cortex with transcranial magnetic stimulation (TMS) based on the functional relation between induced electric fields at the cortical level and peripheral muscle activation (motor-evoked potentials; MEPs). Besides providing insights into structure-function relationships, this routine lays the foundation for TMS dosing metrics based on subject-specific cortical electric field thresholds. MEPs for different coil positions and orientations are combined with electric field modeling, exploiting the causal nature of neuronal activation to pinpoint the cortical origin of the MEPs. This involves constructing an individual head model using magnetic resonance imaging, recording MEPs via electromyography during TMS and computing the induced electric fields with numerical modeling. The cortical muscle representations are determined by relating the TMS-induced electric fields to the MEP amplitudes. Subsequently, the coil position to optimally stimulate the origin of the identified cortical MEP can be determined by numerical modeling. The protocol requires 2 h of manual preparation, 10 h for the automated head model construction, one TMS session lasting 2 h, 12 h of computational postprocessing and an optional second TMS session lasting 30 min. A basic level of computer science expertise and standard TMS neuronavigation equipment suffices to perform the protocol.
https://doi.org/10.1038/s41596-022-00776-6
Attitude reconstruction of a spacecraft from temperature measurements in solar eclipse analysis and observer design for a not globally observable non-linear system. - In: IEEE transactions on control systems technology, ISSN 1558-0865, Bd. 31 (2023), 2, S. 631-645
This article proposes a method that uses only a single temperature measurement and angular velocity measurements to estimate the attitude of a spacecraft under the influence of solely infrared irradiation. The system governing the dynamics is highly non-linear with its attitude being defined in the quaternion space. The resulting observability mapping is used to transform the system into a canonical observability form. However, this mapping is not bijective, and a method to find the arising local inverses is proposed. The reconstruction algorithm itself is divided into two separate parts and uses the canonical form. The first part carries out the dynamic estimation of the temperature and its derivatives. The second part uses these derivatives to estimate the attitude by solving a system of non-linear equations. The proposed algorithm achieves the desired results under the assumption that a suitable initial guess of the attitude is available. Numerical simulations show the validity of the algorithm and illustrate errors induced by measurement noise.
https://doi.org/10.1109/TCST.2022.3187916
Analysis of sliding-mode control systems with relative degree altering perturbations. - In: Automatica, ISSN 0005-1098, Bd. 148 (2023), 110745
We consider sliding-mode control systems subject to unmatched perturbations. Classical first-order sliding-mode techniques are capable to compensate unmatched perturbations if differentiations of the output of sufficiently high order are included in the sliding variable. For such perturbations it is commonly assumed that they do not affect the relative degree of the system. In this contribution we consider perturbations that alter the relative degree of the process and study their impact on the closed-loop control system with a classical first-order sliding-mode design. In particular we consider systems with full (nominal) relative degree subject to a perturbation reducing the relative degree by one and analyse the resulting closed-loop system. It turns out that the sliding-manifold is not of reduced dimension and the uniqueness of the solution may be lost. Also attractivity of the sliding-manifold and global stability of the origin may be lost whereas the disturbance rejection properties of the sliding-mode control are not impaired. We present a necessary and sufficient condition for the existence of unique solutions for the closed-loop system. The second-order case is studied in great detail and allows to parametrically specify the conditions obtained before. We derive a necessary condition for the global asymptotic stability of the closed-loop system. Further we present a constructive condition for the global asymptotic stability of the closed-loop system using a piece-wise linear Lyapunov function. Each of the prominent results is illustrated by a numerical example.
https://doi.org/10.1016/j.automatica.2022.110745
Alternating complexity of counting first-order logic for the subword order. - In: Acta informatica, ISSN 1432-0525, Bd. 60 (2023), 1, S. 79-100
This paper considers the structure consisting of the set of all words over a given alphabet together with the subword relation, regular predicates, and constants for every word. We are interested in the counting extension of first-order logic by threshold counting quantifiers. The main result shows that the two-variable fragment of this logic can be decided in twofold exponential alternating time with linearly many alternations (and therefore in particular in twofold exponential space as announced in the conference version (Kuske and Schwarz, in: MFCS’20, Leibniz International Proceedings in Informatics (LIPIcs) vol. 170, pp 56:1-56:13. Schloss Dagstuhl - Leibniz-Zentrum für Informatik, 2020) of this paper) provided the regular predicates are restricted to piecewise testable ones. This result improves prior insights by Karandikar and Schnoebelen by extending the logic and saving one exponent in the space bound. Its proof consists of two main parts: First, we provide a quantifier elimination procedure that results in a formula with constants of bounded length (this generalises the procedure by Karandikar and Schnoebelen for first-order logic). From this, it follows that quantification in formulas can be restricted to words of bounded length, i.e., the second part of the proof is an adaptation of the method by Ferrante and Rackoff to counting logic and deviates significantly from the path of reasoning by Karandikar and Schnoebelen.
https://doi.org/10.1007/s00236-022-00424-2
Improved ultrasonic sensing using machine learning. - In: AmE 2022, (2022), S. 25-26
We present a novel approach for using industrial grade ultrasonic sensors to perform echolocation by detecting ultrasonic echoes in a noisy environment using machine learning. Autonomous driving is expected to become a huge market and among other technical challenges, environmental perception will be the most critical one. For high automation level, classical technologies are limited. On the other hand, automotive is cost sensitive. The main part of the lecture starts with state of the art technology and then explains how we have used machine learning approaches to train a net for several classifications tasks: Distinguish whether the ultrasonic echo comes from the sensor or another noise source, distinguish whether the echo is relevant or not and finally a height classification. Results are presented in the form of F1-Score. In addition to this, a method will be presented to use CNN for noise suppression in real time. We demonstrate the potential of using the “bat principle” for perception and prove that by that we also achieve the low-cost targets.
https://ieeexplore.ieee.org/document/10025906
An OCL implementation for model-driven engineering of C++. - In: Software engineering research, management and applications, (2022), S. 151-168
Unlike traditional development techniques, Model-Driven Software Development utilizes models as the cornerstone of a software development process and the basis for automated generation of required development artifacts. Its goal is to automate transformations between models and source code. The Object Constraint Language (OCL) is a standard method for querying and validating standardized UML or Ecore models. Several toolchains implement this approach, for example the common Eclipse Modeling Framework (EMF). However, most of these solutions are based on Java or are proprietary solutions. A reason for this is that open source implementations for C++ based on an explicit standardized meta-model are still missing, which are necessary to query a model using OCL during runtime. The Model-Driven Engineering for C++ (MDE4CPP) project is an EMF-like, model-driven environment for common Eclipse Ecore and several OMG specifications like UML, fUML, or PSCS. Although already supporting the execution of UML models, the project did not support OCL so far. This paper presents the concept and implementation of OCL4CPP: an OCL parsing tool for checking, querying, and validating Ecore and UML models at run time within MDE4CPP. It describes implementation details as well as the use of the OCL parser for example applications.
https://doi.org/10.1007/978-3-031-09145-2_10
"Immer vor Ort - mobile medizintechnische Lösungen für eine patientenfreundliche Gesundheitsversorgung" : Sachbericht zum Verwendungsnachweis : zum Verbundprojekt: Plenophthalmologische Kamera für die mobile 3D-Netzhautdiagnostik : Akronym: PlenoM : Berichtszeitraum: 01.01.2019-30.06.2022. - [Ilmenau] : [Technische Universität Ilmenau, Institut für Biomedizinische Technik und Informatik - Fachgebiet Optoelektrophysiologische Medizintechnik]. - 1 Online-Ressource (19 Seiten, 1,83 MB)Literaturverzeichnis: Blatt 3
https://edocs.tib.eu/files/e01fb24/1878364243.pdf
Accelerating Python UDFs in vectorized query execution. - [USA?] : CIDR Conference. - 1 Online-Ressource (7 Seiten)Publikation entstand im Rahmen der Veranstaltung: CIDR 2022 : 12th Annual Conference on Innovative Data Systems Research (CIDR ’22), January 9-12, 2022, Chaminade, USA
https://doi.org/10.22032/dbt.59388
Rate and oscillatory switching dynamics of a multilayer visual microcircuit model. - In: eLife, ISSN 2050-084X, Bd. 11 (2022), e77594, S. 1-28, insges. 28 S.
The neocortex is organized around layered microcircuits consisting of a variety of excitatory and inhibitory neuronal types which perform rate- and oscillation-based computations. Using modeling, we show that both superficial and deep layers of the primary mouse visual cortex implement two ultrasensitive and bistable switches built on mutual inhibitory connectivity motives between somatostatin, parvalbumin, and vasoactive intestinal polypeptide cells. The switches toggle pyramidal neurons between high and low firing rate states that are synchronized across layers through translaminar connectivity. Moreover, inhibited and disinhibited states are characterized by low- and high-frequency oscillations, respectively, with layer-specific differences in frequency and power which show asymmetric changes during state transitions. These findings are consistent with a number of experimental observations and embed firing rate together with oscillatory changes within a switch interpretation of the microcircuit.
https://doi.org/10.7554/eLife.77594