Daniel Hsu |
**Information for prospective students**

Please see the CS department's webpage on admissions. There, you'll find information about all aspects of the application process (e.g., deadlines, your chances of being admitted), as well as information about the academic programs themselves.

You may also be interested in academic opportunities through the Institute for Data Sciences and Engineering, as well as the NSF IGERT program "From Data to Solutions".

If you are interested in working with me, you can mention this in
your application (and also see the section below); this ensures
that I (and others) will review your file. **Unfortunately, I do
not generally have time to reply to inquiries from prospective
students.**

I work on algorithmic statistics and machine learning, with emphases on

- probabilistic models (especially latent variable models),
- interactive learning, and
- privacy-preserving data analysis.

- Mixture models: wikipedia page, a theoretical computer science perspective by Kamalika Chaudhuri
- Topic models: wikipedia page
- Active learning: survey paper by Sanjoy Dasgupta
- Contextual bandits: talk by Lev Reyzin on vimeo
- Differential privacy: survey paper by Cynthia Dwork, some talk slides

I am also generally interested in algorithms for unsupervised and semi-supervised machine learning.

Most of my research is theoretical in nature, although I am also very interested in some practical applications (e.g., in natural language processing, computational biology, information retrieval). So the actual work involved in my research is quite varied.

The mathematical background required for research in the above topics includes: algorithms, probability theory, calculus, linear algebra, and convex analysis. Solid programming skills are essential for good experimental work; fluency in C/C++ is helpful for scalable implementations, although prototyping can be done in higher-level environments such as MATLAB and Python.

If you are a current (or already admitted) Columbia student, please feel free to contact me via e-mail (or drop by my office) if you are interested in discussing a project with me.