Assistant Professor, Statistics
Institute of Science and Technology (IST) Austria
Am Campus 1
3400 Klosterneuburg, Austria
email: caroline.uhler [@] ist.ac.at
Research area in the programme
Convex and Stochastic Optimization with Applications in Biology
- Convex optimization
- convex relaxations
- convex geometry
- structure and parameter estimation in graphical models
- high-dimensional statistics
- sparsity and robustness in graphical models
- statistical modelling in biology
Caroline Uhler works at the intersection of mathematical statistics, optimization and biomathematics. Her research is motivated by questions from biological applications such as: How are chromosomes packed into the cell nucleus? How many observations are minimally needed for estimating interactions between genes? How can privacy be ensured when releasing genomic data? On the theoretical side she is mainly interested in developing new methods for causal inference and hypothesis testing in the setting where the number of variables exceeds the number of observations. On the applied side, Caroline Uhler is intrigued by the possible applications of stochastic and convex optimization techniques to model and understand the emergence of different geometric patterns in biology. Related projects include studying the consequences of nucleus shape on the arrangement of chromosomes and the changes in chromosome arrangement in cancer development, the geometric clustering of receptors on the cell membrane, and the packing of neurons in the brain. All these problems require modeling and then solving large-scale optimization problems.
Know-how and infrastructure of the research group
Caroline Uhler’s expertise is in the area of graphical models, algebraic statistics and convex optimization. She has done substantial work in structure and parameter estimation in graphical models using tools from convex geometry and computational algebra. A special emphasis of Caroline’s research is on causal inference using graphical models. In a recent paper she has shown that the most widely used algorithm, the prominent PC algorithm, has severe limitations and should not be used for causal inference. Due to her biological interests she also has various ongoing collaborations with experimental biologists. She has developed an algorithm for finding minimal overlap configurations of ellipsoids inside an ellipsoidal container, which can be used as a basis to study the emergence of different geometric patterns in biology.
The group of Caroline Uhler currently consists of three postdocs and two PhD students. The group members have a diverse background (mathematical statistics, applied algebraic geometry, algebraic combinatorics, computational geometry and computational biology) creating a stimulating and resourceful research environment. The infrastructure (e.g. well-equipped offices, computing facilities) necessary to perform the proposed projects is provided by IST Austria.
Assistant professor at IST Austria since 11/2011. Postdoctoral Researcher in the Seminar for Statistics at ETH Zurich from 01/2012 to 07/2012. Simons Fellow in the program "Theoretical Foundations of Big Data Analysis" at the Simons Institute from 08/2013-01/2014.
B.Sc. (2004) and M.Sc. (2006) in Mathematics with a Minor in Biology from the University of Zurich. M.Ed. in High School Mathematical Education from the University of Zurich (2007). Ph.D. in Statistics with a Designated Emphasis in Computational and Genomic Biology from UC Berkeley (2011). Management of Technology Degree from UC Berkeley (2011).
Best Student Award from the Universiy of Zurich (2011). International Fulbright Science and Technology Award (full tuition and stipend 2007-2010, 30 fellows per year worldwide). Janggen-Poehn Fellowship (full tuition and stipend 2010-2011). Golden Chalk Award for excellence in teaching from IST Austria (2013). Research Fellowship from the Simons Institute at UC Berkeley for the program "Theoretical Foundations of Big Data Analysis" during Fall 2013.
Collaborations within the programme
Georg Pflug on statistical data analysis and in particular on causal inference
Immanuel Bomze on semidefinite programming and applications to the natural sciences
University of Vienna
1090 Vienna, Austria
T: +43-1-4277-386 31