Disputation: Konstantinos Nikolaidis

Doctoral candidate Konstantinos Nikolaidis at the Department of informatics, Faculty of Mathematics and Natural Sciences, is defending the thesis Improving the Classification of Time-Series Data for Sleep Apnea Detection for the degree of Philosophiae Doctor.

Picture of the candidate

Photo: Private

The PhD defence and trial lecture will be fully digital and streamed directly using Zoom. The host of the session will moderate the technicalities while the chair of the defence will moderate the disputation.

Ex auditorio questions: the chair of the defence will invite the audience to ask ex auditorio questions either written or oral. This can be requested by clicking 'Participants -> Raise hand'. 

Trial lecture

Title:

"Domain adaptation and zero-shot learning - methods and applications".

 

Main research findings

  • Sleep apnea (SA) is a severe, common, and largely under-diagnosed
    sleep-related breathing disorder. In this PhD Thesis, we aim to reliably
    perform SA detection with low cost sensors and Machine Learning (ML). We train ML models with many examples (i.e., supervised learning) in form
    of sleep monitoring data that is labeled by sleep experts. In the early
    phases of the project we could demonstrate that ML is able to detect SA
    with high performance, even if only a single signal is used. However,
    this works only if there is enough training data with sufficient
    quality. This is often not given for SA, because (1) most sleep
    monitoring data cannot be shared due to privacy concerns, (2) the need
    for sleep experts makes labelling expensive and can introduce label
    errors, and (3) it is unclear whether different sensors for the same
    signal represent different domains. To address these issues we developed
    in this PhD Thesis new solutions to augment sleep monitoring data sets
    with synthetic data, to generate anonymous data sets that are robust
    against known attacks, to correct wrong labels, and to enable adaptation
    of ML models to new domains (i.e., new sensors).

Adjudication committee:

  • Professor Shin'ichi Satoh, National Institute of Informatics (NII), Tokyo, Japan
  • Associate Professor Carolina Varon Perez, Technische Universiteit Delft, The Netherlands
  • Professor Torbjørn Rognes, Department of Informatics, University of Oslo, Norway

Supervisors

  • Professor Vera Goebel, Department of Informatics, UiO
  • Researcher Stein Kristiansen, Department of Informatics, UiO
  • Professor Knut Liestøl, Department of Informatics, UiO
  • Professor Mohan Kankanhalli, National University of Singapore

Chair of defence:

Associate Professor Petter Nielsen, Department of Informatics, UiO

Contact information to Department: Mozhdeh Sheibani Harat

Publisert 1. sep. 2021 13:05 - Sist endret 15. sep. 2021 14:49