Taygun Kekeç

Postdoc at Biomedical Imaging Group
Department of Radiology, Erasmus Medical Centre
Doctor Molewaterplein 40, 3015 GD Rotterdam
Email: namesurname at gmail c-o-m


He is currently a postdoc researcher in the Biomedical Imaging Group (BIGR) at Erasmus MC. He works in the project "CTA Image Analysis For Acute Stroke" which targets to find biomarkers that can then be used to provide guidelines for the stroke treatment outcome, and assist the clinical decision-making process. He completed his PhD, Exponential Word Embeddings: Models and Approximation, in the Pattern Recognition and Bioinformatics (PRB) Laboratory, under the supervision of David M. J. Tax at Delft University of Technology. In his PhD thesis; he investigated various unsupervised learning models for obtaining word representations. These representations can be used in many language processing tasks when there is a need of mapping words to feature vectors.

His research is centralized around the following question: "Given an agent A, what are the limitations of unsupervised pretraining for task T?" His PhD was an application of this issue for the word embedding problems with the following learning settings: noisy data, access to negative samples, and large capacity model. He currently analyzes limitations of pretraining in medical images.