Samory Kpotufe

E-mail: samory@<institute domain>
I am a researcher at the Max Planck Institute for Intelligent Systems.
Office: SpemannStrasse 38, 214

I work in machine learning, and I'm particularly interested in nonparametric methods. My research is concerned with characteristics of high-dimensional data which can help improve learning. Some such characteristics are the intrinsic dimension of data (e.g. sparse data or data on a manifold), or the correlation between labels and clusters in the data.
One of the main questions on my mind, is whether dimension reduction is necessary when dealing with high-dimensional data. The partial answer, as shown by recent results in the field, is that some learners perform well on high-dimensional data without dimension reduction. These learners converge at rates that depend only on the intrinsic dimension of data (e.g. manifold dimension). A more complete answer to this question stands to impact on how we deal with high-dimensional data in practice.

For a more in depth discussion of this question, take a look at my research statement.
Also, here is an updated talk I've been giving on the subject of adaptivity in nonparametric regression.

I am in the Empirical Inference Department of Prof. Dr. Bernhard Schölkopf,
working mainly with the Learning Theory group of Dr. Ulrike von Luxburg.
My Ph.D. advisor was Prof. Sanjoy Dasgupta at UCSD CSE.

Research

Samory Kpotufe. k-NN Regression adapts to local intrinsic dimension.
Neural Information Processing Sytems (NIPS) 2011 (accepted as a talk) [ pdf]

Samory Kpotufe, Ulrike von Luxburg. Pruning nearest neighbor cluster trees.
International Conference on Machine Learning (ICML) 2011. [ pdf | slides ]

Samory Kpotufe, Sanjoy Dasgupta. A tree-based regressor that adapts to intrinsic dimension.
(To appear) Special Issue of the Journal of Computer and Systems Sciences 2011. [ pdf ]

Samory Kpotufe. The curse of dimension in nonparametric regression.
UCSD, Phd Dissertation 2010. [ pdf ]

Samory Kpotufe. Escaping the curse of dimensionality with a tree-based regressor.
Conference on Learning Theory (COLT) 2009. Mark Fulk Best Student Paper at COLT 09. [ pdf | slides ]

Nakul Verma, Samory Kpotufe, Sanjoy Dasgupta. Which spatial partition trees are adaptive to intrinsic dimension?
Uncertainty in Artificial Intelligence (UAI) 2009. [ pdf | poster ]

Samory Kpotufe. Fast, smooth and adaptive regression in metric spaces.
Neural Information Processing Sytems (NIPS) 2009. [ pdf ]

Invited Talks

Weierstrass Institute for Applied Analysis and Stochastics. November 2011.

Foundations of Computational Mathematics, Learning Theory Workshop. June 2011.

University College London, Gatsby Unit. March 2011.

University of Stuttgart, Institute for Stochastics and Applications. November 2010.

Los Alamos National Lab, Engineering Institute. August 2009.

Professional activities

Reviewed for Journal of Machine Learning Research, IEEE Transactions on Pattern Analysis and Machine Intelligence, Neural Information Processing Systems, ACM-SIAM Symposium On Discrete Algorithms.

Education

Phd in Computer Science and Engineering, UC San Diego, 2010.

M.S. in Computer Science and Engineering, UC San Diego, 2007.

B.S. in Mathematics, and B.S. in Computer Science, University of Denver, 1999.

Teaching assistanship

CSE 20 - Discrete Mathematics

CSE 101 - Design and Analysis of Algorithms