Universität Stuttgart

Permanent URI for this communityhttps://elib.uni-stuttgart.de/handle/11682/1

Browse

Filters

2021 - 20233

Settings

Institute: search.filters.UBSInstitut.Institut für SystemdynamikSubject: search.filters.subject.004

Search Results

Now showing 1 - 3 of 3
  • Thumbnail Image
    ItemOpen Access
    FeaSel-Net : a recursive feature selection callback in neural networks
    (2022) Fischer, Felix; Birk, Alexander; Somers, Peter; Frenner, Karsten; Tarín, Cristina; Herkommer, Alois
    Selecting only the relevant subsets from all gathered data has never been as challenging as it is in these times of big data and sensor fusion. Multiple complementary methods have emerged for the observation of similar phenomena; oftentimes, many of these techniques are superimposed in order to make the best possible decisions. A pathologist, for example, uses microscopic and spectroscopic techniques to discriminate between healthy and cancerous tissue. Especially in the field of spectroscopy in medicine, an immense number of frequencies are recorded and appropriately sized datasets are rarely acquired due to the time-intensive measurements and the lack of patients. In order to cope with the curse of dimensionality in machine learning, it is necessary to reduce the overhead from irrelevant or redundant features. In this article, we propose a feature selection callback algorithm (FeaSel-Net) that can be embedded in deep neural networks. It recursively prunes the input nodes after the optimizer in the neural network achieves satisfying results. We demonstrate the performance of the feature selection algorithm on different publicly available datasets and compare it to existing feature selection methods. Our algorithm combines the advantages of neural networks’ nonlinear learning ability and the embedding of the feature selection algorithm into the actual classifier optimization.
  • Thumbnail Image
    ItemOpen Access
    Machine learning methods of regression for plasmonic nanoantenna glucose sensing
    (2021) Corcione, Emilio; Pfezer, Diana; Hentschel, Mario; Giessen, Harald; Tarín, Cristina
    The measurement and quantification of glucose concentrations is a field of major interest, whether motivated by potential clinical applications or as a prime example of biosensing in basic research. In recent years, optical sensing methods have emerged as promising glucose measurement techniques in the literature, with surface-enhanced infrared absorption (SEIRA) spectroscopy combining the sensitivity of plasmonic systems and the specificity of standard infrared spectroscopy. The challenge addressed in this paper is to determine the best method to estimate the glucose concentration in aqueous solutions in the presence of fructose from the measured reflectance spectra. This is referred to as the inverse problem of sensing and usually solved via linear regression. Here, instead, several advanced machine learning regression algorithms are proposed and compared, while the sensor data are subject to a pre-processing routine aiming to isolate key patterns from which to extract the relevant information. The most accurate and reliable predictions were finally made by a Gaussian process regression model which improves by more than 60% on previous approaches. Our findings give insight into the applicability of machine learning methods of regression for sensor calibration and explore the limitations of SEIRA glucose sensing.
  • Thumbnail Image
    ItemOpen Access
    Cystoscopic depth estimation using gated adversarial domain adaptation
    (2023) Somers, Peter; Holdenried-Krafft, Simon; Zahn, Johannes; Schüle, Johannes; Veil, Carina; Harland, Niklas; Walz, Simon; Stenzl, Arnulf; Sawodny, Oliver; Tarín, Cristina; Lensch, Hendrik P. A.
    Monocular depth estimation from camera images is very important for surrounding scene evaluation in many technical fields from automotive to medicine. However, traditional triangulation methods using stereo cameras or multiple views with the assumption of a rigid environment are not applicable for endoscopic domains. Particularly in cystoscopies it is not possible to produce ground truth depth information to directly train machine learning algorithms for using a monocular image directly for depth prediction. This work considers first creating a synthetic cystoscopic environment for initial encoding of depth information from synthetically rendered images. Next, the task of predicting pixel-wise depth values for real images is constrained to a domain adaption between the synthetic and real image domains. This adaptation is done through added gated residual blocks in order to simplify the network task and maintain training stability during adversarial training. Training is done on an internally collected cystoscopy dataset from human patients. The results after training demonstrate the ability to predict reasonable depth estimations from actual cystoscopic videos and added stability from using gated residual blocks is shown to prevent mode collapse during adversarial training.