Hsiang Hsu

  • I am a third year Ph.D. candidate in Computer Science department at Harvard University, working with Professor Flavio P. Calmon in Learning, Information and Knowledge Lab.
  • My current research focuses on representation learning and robustness of machine learning, in theory and in practice. Most recently, I am working on deep orthogonal representations on images for correspondence analysis and multi-view learning, and f-information bottleneck.
  • I received M.S. in Electrical Engineering in 2016, and B.S. in Mathematics and Electrical Engineering in 2014, both from National Taiwan University (NTU). Previously, I worked with Professor Kwang-Cheng Chen on crowdsourcing and wireless caching, and with Professor Huan-Cheng Chang on fluorescent nano-diamonds and micro-imaging.


Nov 11, 2019: Our paper Discovering Information-Leaking Samples and Features is selected as spotlight talk in NeurIPS PriML 2019. [Paper][Slides][Poster]

Apr 1, 2019: Our paper Information-Theoretic Privacy Watchdogs is accepted by ISIT 2019. [Paper][Slides]

Dec 22, 2018: Our paper Correspondence Analysis Using Neural Networks is accepted by AISTATS 2019. [Paper][Poster]

Dec 7, 2018: Our paper Correspondence Analysis of Government Expenditure Patterns will be presented in 2019 NeurIPS Workshops. [Paper][Poster]

Jun 21, 2018: Check out our latest work on Deep Orthogonal Representations: Fundamental Properties and Applications. [Paper]

May 7, 2018: I am invited to New England Machine Learning (NEML) Day at Microsoft Research. [Poster]

Apr 3, 2018: Our paper Generalizing Bottleneck Problems is accepted by ISIT 2018. [Paper][Slides]


Email: hsianghsu(at)g(dot)harvard(dot)edu
Office: 33 Oxford St., Room 321, Cambridge, MA 02138