David Purdy




I am a Ph.D. candidate in Statistics, in Bin Yu's group.  My interests include machine learning, the analysis and modeling of very high dimensional data (~1 million dimensions, comparable observations), and applications in natural language processing and information retrieval. I've also enjoyed working in quantitative finance. Because building the best models requires careful understanding of the data and the models, I'm also very interested in visualization of the relationship between data and models, and across multiple models.

My primary research projects include: ranking for very high dimensional text transliteration data (e.g. Romanji to Japanese or Pinyin to Chinese), interactive visualization of one or multiple models, and algorithmic identification and prediction of errors by models, known as calibration.


Activities and affiliations

  • Statistics Graduate Students Association (wiki) I was the SGSA President for 2007-2008. The new President is Megan Goldman

  • I established a site, Statistics Careers, which is now the main departmental channel for distributing relevant information on jobs and events.  Access is open to all members of the Cal community, via your bSpace/CalNet login.  Employers may use this link for announcing job opportunities.

  • I am on the Graduate Student Advisory Group for the Career Center.

  • Statistical consulting: I've done several short-term projects.  Several stimulating projects related to information retrieval, spam, computer security and operating system bugs, auction forecasting, models for ranking medical devices, and collaborative filtering.  I also recommend the department's consulting services.  I'm open to discuss areas that I know something about; if I'm too constrained on time, I may be able to refer you to someone else.

Selected past activities and affiliations

  • UC Berkeley Graduate Assembly: I was the SGSA delegate for two academic years.

  • VIGRE Seminar Series: I co-coordinated this series for undergraduates, for two years. Interesting speakers are always sought!

  • Microsoft Research Asia, Beijing: Where I began working with the IME data set.

  • Yahoo and AltaVista: I worked in web search research, on applications of machine learning and statistics to problems such as search results ranking, web page summarization, query classification, among other areas.

  • The U.S. National Security Agency: My work at the NSA involved cryptanalysis, pure mathematics, and statistical machine translation.  In addition, the NSA sponsored my undergraduate education through what is now called the Stokes Educational Scholarship Program.

  • Caltech  and Harvard, for my undergraduate studies in math (my major), along with many courses in linguistics, Japanese, and EECS topics.  Noam Elkies had the good humor to be my advisor for my senior thesis on computational algorithms for searching for high rank elliptic curves.


I'd like to send you to:

Email username: dpurdy
Email server: stat.berkeley.edu