Severin Gsponer

PhD Student
The Insight Centre for Data Analytics
University College Dublin

Contact

severin.gsponer@insight-centre.org

Short bio

Since 2015 Severin Gsponer is a PhD student at the Insight Centre for Data Analytics at the University College Dublin, Ireland. He holds a MSc (2014, University of Basel, Switzerland) and a BSc (2012, University of Basel, Switzerland) in Computer Science.
His current research focuses on sequence regression, in particular the analysis of different loss functions and convex optimization strategies within the sequence learning framework SEQL. In this context, he is especially interested in the scalability as well as the interpretability of the produced models. Additionally he works on improvements regarding feature selection strategies in huge feature spaces with structure, such as n-gram spaces.

Main Research Interests

  • Machine learning
  • Optimization
  • Sequence Learning
  • Evolutionary computing
  • Reproducibility in research

Talks

  • Efficient Sequence Regression by Learning Linear Models in All-Subsequence Space. Slides September 2017
  • Introduction to Git for Master Students. Slides October 2016
  • Machine Learning for Sequence Learning. Slides January 2016

Teaching

Teaching assistant / Tutor

  • Object-Oriented Programming (COMP30070) WS 2016/ WS 2017
  • Software Engineering (COMP30670) SS 2016
  • Computer Programming I (COMP10110) WS 2015
  • Introduction to Java (COMP20250) WS 2015
  • Programming I (Python) (COMP10280) WS 2015

Publications

  • Efficient Sequence Regression by Learning Linear Models in All-Subsequence Space, ECML-PKDD 2017, Skopje, Macedonia, Severin Gsponer, Barry Smyth, Georgiana Ifrim (Insight Centre for Data Analytics - University College Dublin). <pdf> <code>
  • Time Series Classification by Sequence Learning in All-Subsequence Space, ICDE 2017 IEEE International Conference on Data Engineering, San Diego, Thach Le Nguyen, Severin Gsponer, Georgiana Ifrim (Insight Centre for Data Analytics - University College Dublin). <pdf> <code>
  • Reproducible Experiments in High-Performance Computing: Techniques and Stencil Compiler Benchmark Study, Presentation at Platform for Advanced Scientific Computing 2014 , ETH Zürich, Helmar Burkhart, Severin Gsponer, Danilo Guerrera, and Antonio Maffia (University of Basel).
  • Reproducible Experiments in High-Performance Computing: Techniques and Stencil Compiler Benchmark Study, Workshop at 1st International Workshop on Reproducibility in Parallel Computing, Porto, Portugal, Danilo Guerrera, Helmar Burkhart, Severin Gsponer, and Antonio Maffia (University of Basel).

Smart Pointers

Podcasts

English podcasts

  • Linear Digressions - Regular podcast by Ben Jaffe and Katie Malone about various topics in Machine learning.
  • Data Sceptic - Podcast with Kyle and Linh Da features longer interviews with active data scientists as well as episodes that introduces basic principles of Data Science.
  • Talking Machines - Katherine Gorman and Neil Lawrence talk about topics in Machine learning and present interesting interviews with researchers and experts in the field.

German podcasts

  • Methodisch inkorrekt - Nicolas Wöhrl and Reinhard Remfort present fortnightly four papers from fields of science in a entertaining and informative way.

Newsletters

These newsletters provide a good overview over the mass of article and news in the field of ML and AI:

  • ML Weekly - Gathering of ML related Twitter, Facebook, Google+, and LinkedIn posts.
  • AI Weekly - Similar to ML Weekly but for the field of AI.
  • Data Elixir - Curated news and resources about data science.
  • Import AI - A newsletter about artificial intelligence.

These mailing lists contain recent announcements and CfP and are probably more interesting for researchers and people working in academia.

  • Connectionists - Mailing list about neural networks and cognitive or computational neuroscience.
  • ML-News - Google group for announcements and CfP for the machine learning community.
  • UAI - Mailing list for uncertainty in artificial intelligence.