[Forschungsseminar-BSV] Forschungsseminar Computergrafik, Bildverarbeitung und Visualisierung

Vanessa Kretzschmar kretzschmar at informatik.uni-leipzig.de
Do Jun 2 13:25:10 CEST 2022


Achtung:
  - mehrere Vortraege (erhoehte Dauer moeglich)


E I N L A D U N G

======================================================================

zum Forschungsseminar 'Computergrafik, Bildverarbeitung und Visualisierung'

     am Mittwoch, den 8. Juni 2022, 13:15 Uhr,
     im Raum P-701 im Paulinum am Augustusplatz, sowie über eine 
Webkonferenz.
     (https://conf.fmi.uni-leipzig.de/b/van-bwb-jil-vk2)

======================================================================

Wir hören einen Vortrag von

     Christofer Meinecke

mit dem Titel:

     ' Explaining Semi-Supervised Text Alignment through Visualization'

zum Inhalt:

     The analysis of variance in complex text traditions is an arduous task
     when carried out manually. Text alignment algorithms provide domain
     experts with a robust alternative to such repetitive tasks. Existing
     white-box approaches allow the digital humanities to establish 
syntax-based
     metrics taking into account the spelling, morphology and order of 
words.
     However, they produce limited results, as semantic meanings are 
typically
     not taken into account. Our interdisciplinary collaboration between
     visualization and digital humanities combined a semi-supervised text
     alignment approach based on word embeddings that take not only 
syntactic
     but also semantic text features into account, thereby improving the
     overall quality of the alignment. In our collaboration, we developed
     different visual interfaces that communicate the word distribution in
     high-dimensional vector space generated by the underlying neural 
network
     for increased transparency, assessment of the tools reliability and
     overall improved hypothesis generation. We further offer visual means
     to enable the expert reader to feed domain knowledge into the system at
     multiple levels with the aim of improving both the product and the 
process
     of text alignment. This ultimately illustrates how visualization can
     engage with and augment complex modes of reading in the humanities.

======================================================================

Weiterhin, hören wir einen Vortrag von

     Leo Sperling (digital zugeschalten)

mit dem Titel:

     'Uncertainty-aware Evaluation of Machine Learning Performance in 
binary Classification Tasks'

zum Inhalt:

     Machine learning has become a standard tool in computer vision. 
Nowadays,
     neural networks are one of the most prominent representatives in this
     class of algorithms that usually require training and evaluation to 
work
     as desired. There exist a variety of evaluation metrics to 
determine the
     quality of a trained neural network, which are usually threshold 
dependent.
     This results in massive changes in the resulting evaluation when the
     threshold is changed slightly. Further, measurements of uncertainty 
such
     as resulting from Bayesian approaches, are not considered in this 
analysis.
     In this paper, we present evaluation metrics for machine learning
     approaches that are able to attach a probability distribution to the
     utilized threshold and include uncertainty measures. We demonstrate the
     applicability of our approach by applying the defined metrics to a
     real-world example where a Bayesian neural network has been used to
     predict stroke lesions

======================================================================

Alle Interessierten sind im Namen von Professor Dr. Scheuermann herzlich
eingeladen.

Mit freundlichen Grüßen
Vanessa Kretzschmar
-------------- nächster Teil --------------
Ein Dateianhang mit Binärdaten wurde abgetrennt...
Dateiname   : 20220608T131500_bsv_invitation.ics
Dateityp    : text/calendar
Dateigröße  : 3612 bytes
Beschreibung: nicht verfügbar
URL         : <http://lists.informatik.uni-leipzig.de/pipermail/forschungsseminar-bsv/attachments/20220602/97058d8d/attachment.ics>


Mehr Informationen über die Mailingliste Forschungsseminar-BSV