Clemens Hutter


Last Name

Hutter

First Name

Clemens

ORCID

0000-0003-3008-8106

Organisational unit

03610 - Boelcskei, Helmut / Boelcskei, Helmut

Filters

2021 - 20254
Reset filters

Search Results

Publications 1 - 4 of 4
  • Recurrent neural network approximation theory
    Item type: Doctoral Thesis
    Hutter, Clemens (2025)
    In this thesis we study the approximation capabilities of recurrent neural networks (RNNs). Firstly, we consider the approximation of certain classes of systems mapping an input sequence to an output sequence. We prove that RNNs can approximate any such system to within arbitrarily small worst-case error. What is more, we derive quantitative results on the amount of information needed to specify the approximating RNN. Furthermore, we present a framework to unify the study of different approximation tasks, allowing us to conclude that neural network learning is universally information-optimal for a large variety of approximation problems throughout both function and system approximation. Secondly, we study the use of RNNs for approximating real-valued polynomials. We find that for every polynomial there exists an RNN that produces — as its output sequence — increasingly precise approximations to the target polynomial. This is remarkable as the weights of the RNN do not depend on the desired approximation error. That is, any arbitrarily small approximation error is achieved simply by running the approximating RNN for more time steps.
  • Profiling Daily Life Performance Recovery in the Early Subacute Phase After Stroke Using a Graphical Modeling Approach
    Item type: Journal Article
    Veerbeek, Janne M.; Hutter, Clemens; Ottiger, Beatrice; et al. (2023)
    Journal of the American Heart Association
    Background Laboratory-based assessments have shown that stroke recovery is heterogeneous between patients and affected domains such as motor and language function. However, laboratory-based assessments are not ecologically valid and do not necessarily reflect patients' daily life performance. Therefore, we aimed to give an innovative view on stroke recovery by profiling daily life performance recovery across domains in patients with early subacute stroke and determine their interrelatedness, taking stroke localization into account. Methods and Results Daily life performance was observed at neurorehabilitation admission and weekly thereafter until discharge, using a scale containing 7 daily life domains. Graphical modeling was applied to investigate the conditional independence between recovery of these domains depending on stroke localization. There were 592 patients analyzed. Four clusters of interrelated domains were identified within the first 6 weeks poststroke. The first cluster included recovery in learning and applying knowledge, general tasks and demands, and domestic life. The second cluster comprised recovery in self-care and general tasks and demands. The third cluster included recovery in mobility and self-care; it incorporated interpersonal interactions and relationships in left supratentorial stroke, and learning and applying knowledge in right supratentorial stroke. The final cluster included only communication recovery. Conclusions Daily life recovery dynamics early poststroke show that although impairments in body functions are anatomically determined, their impact on performance is comparable. Second, some, but by no means all, domains show an interrelated recovery. Domains requiring cognitive abilities are especially interrelated and seem to be essential for concomitant recovery in mobility and domestic life.
  • Knowledge transfer across cell lines using hybrid Gaussian process models with entity embedding vectors
    Item type: Journal Article
    Hutter, Clemens; von Stosch, Moritz; Cruz Bournazou, Mariano N.; et al. (2021)
    Biotechnology and Bioengineering
    To date, a large number of experiments are performed to develop a biochemical process. The generated data is used only once, to take decisions for development. Could we exploit data of already developed processes to make predictions for a novel process, we could significantly reduce the number of experiments needed. Processes for different products exhibit differences in behaviour, typically only a subset behave similar. Therefore, effective learning on multiple product spanning process data requires a sensible representation of the product identity. We propose to represent the product identity (a categorical feature) by embedding vectors that serve as input to a Gaussian process regression model. We demonstrate how the embedding vectors can be learned from process data and show that they capture an interpretable notion of product similarity. The improvement in performance is compared to traditional one-hot encoding on a simulated cross product learning task. All in all, the proposed method could render possible significant reductions in wet-lab experiments.
  • Metric entropy limits on recurrent neural network learning of linear dynamical systems
    Item type: Journal Article
    Hutter, Clemens; Gül, Recep; Bölcskei, Helmut (2022)
    Applied and Computational Harmonic Analysis
    One of the most influential results in neural network theory is the universal approximation theorem [1–3] which states that continuous functions can be approximated to within arbitrary accuracy by single-hidden-layer feedforward neural networks. The purpose of this paper is to establish a result in this spirit for the approximation of general discrete-time linear dynamical systems—including time-varying systems—by recurrent neural networks (RNNs). For the subclass of linear time-invariant (LTI) systems, we devise a quantitative version of this statement. Specifically, measuring the complexity of the considered class of LTI systems through metric entropy according to [4], we show that RNNs can optimally learn—or identify in system-theory parlance—stable LTI systems. For LTI systems whose input-output relation is characterized through a difference equation, this means that RNNs can learn the difference equation from input-output traces in a metric-entropy optimal manner.
Publications 1 - 4 of 4