Variational auto-encoders with Student’s t-prior

Summary, in English

We propose a new structure for the variational auto-encoders (VAEs) prior, with the weakly informative multivariate Student’s t-distribution. In the proposed model all distribution parameters are trained, thereby allowing for a more robust approximation of the underlying data distribution. We used Fashion-MNIST data in two experiments to compare the proposed VAEs with the standard Gaussian priors. Both experiments showed a better reconstruction of the images with VAEs using Student’s t-prior distribution.

• Author: Najmeh Abiri; Mattias Ohlsson
• Department/s: Computational Biology and Biological Physics - Undergoing reorganization; eSSENCE: The e-Science Collaboration
• Publishing year: 2019
• Language: English
• Publication/Series: ESANN 2019 - Proceedings : The 27th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning
• Links: https://www.elen.ucl.ac.be/Proceedings/esann/esannpdf/es2019-42.pdf
• Document type: Conference paper
• Topic: Computer Science; Other Computer and Information Science
• Conference name: 27th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning
• Conference date: 2019-04-24 - 2019-04-26
• Conference place: Bruges, Belgium
• Status: Published
• Project: Lund University AI Research
• ISBN/ISSN/Other: ISBN: 978-287-587-065-0