This event has passed.

Dimensionality reduction and manifold learning – towards interpretable image analysis

1 February, 2019 @ 3:00 pm - 4:00 pm

Gemma Piella is an Associate Professor at the UPF since 2010. She obtained her degree in Telecommunication Engineering from the Universitat Politècnica  de Catalunya (UPC), Barcelona, Spain, and a Ph.D. in Mathematics, Applied Sciences, from the Universiteit van Amsterdam, The Netherlands. From 2003 to 2004, she was at UPC as a visiting professor. She then stayed at the École Nationale Supérieure des Télécommunications, Paris, as a post-doctoral Marie Curie fellowship  until 2005. Since then she has been at the UPF, Barcelona, first as a visiting professor, then as a Ramón y Cajal researcher and currently as an Associate Professor. Her research is focused on computerised medical image analysis and data integration, including: machine learning and computer vision.

The variation observed in high-dimensional data can be often described by a low number of parameters. Revealing the low-dimensional representation of such high-dimensional data not only leads to a much compact  description but also helps us understand the underlying structure of the data and the process that generated them. Manifold learning techniques target the conversion of data from high to lower dimensional representations  while preserving the intrinsic geometry of the data space. As a result, the new embedding space may simplify further analysis, for example for data mining, pattern recognition or knowledge discovery.  In this talk, I will explain the basics of manifold learning and dimensionality reduction, and how they can be used for an interpretable analysis of images.
+ Google Calendar+ iCal Export