Andrew I. Schein, Ph.D.

Contact Information
LinkedIn Page
Google Scholar Page
Stack Overflow Profile

Related

Philip Schein's Google Scholar Page

Publications

Andrew I. Schein, Johnnie F. Caver, Randale J. Honaker, and Craig H. Martell. Author Attribution Evaluation with Novel Topic Cross-Validation. Appeared in The 2010 International Conference on Knowledge Discovery and Information Retrieval. October 25-28. Valencia, Spain. [.pdf][data]

Grant Gehrke, Craig Martell, Andrew Schein and Pranav Anand. Projecting Away The Class Imbalance Problem in Author Attribution. International Journal of Semantic Computing. Volume: 3, Issue: 3 (September 2009).

Andrew I. Schein and Lyle H. Ungar. Active Learning for Logistic Regression: An Evaluation. Machine Learning. 68:3 2007. Publisher DOI: 10.1007/s10994-007-5017-7. [.ps.gz] [.pdf]

Jinying Chen, Andrew I. Schein, Martha S. Palmer, and Lyle H. Ungar. An Empirical Study of the Behavior of Active Learning for Word Sense Disambiguation. Appeared in the Proceedings of HLT-NAACL 2006. June 5-7, 2006. New York, NY. [.pdf]

Andrew I. Schein. Active Learning for Logistic Regression. Ph.D. Dissertation in Computer and Information Science. The University of Pennsylvania. Defended: April 21, 2005. Supervisor: Lyle H. Ungar. Committee: Andreas Buja, Mark Liberman, Andrew McCallum (external), and Fernando C. N. Pereira (chair). Talk: [.ps.gz] [.pdf] Doc: [.ps.gz] [.pdf]

Automatic term list generation for entity tagging. Ted Sandler, Andrew I. Schein and Lyle H. Ungar. Bioinformatics Advance Access published on October 25, 2005, DOI 10.1093/bioinformatics/bti733.

Seth Kulick, Ann Bies, Mark Liberman, Mark Mandel, Ryan McDonald, Martha Palmer, Andrew Schein, Lyle Ungar, Scott Winters, and Pete White. Integrating Annotation for Biomedical Information Extraction. BioLINK 2004: Linking Biological Literature, Ontologies and Databases, pp. 61-68. May 6, 2004. Boston, MA. [.pdf]

Andrew I. Schein and Lyle H. Ungar. A-Optimality for Active Learning of Logistic Regression Classifiers. The University of Pennsylvania Department of Computer and Information Science Technical Report No. MS-CIS-04-07: [.ps.gz][.pdf].

Andrew I. Schein, S. Ted Sandler and Lyle H. Ungar. Bayesian Example Selection using BaBiES. The University of Pennsylvania Department of Computer and Information Science Technical Report No. MS-CIS-04-08 [.ps.gz][.pdf].

Seth Kulick, Mark Liberman, Martha Palmer, and Andrew Schein. Shallow Semantic Annotation of Biomedical Corpora for Information Extraction. Proceedings of the 2003 ISMB Special Interest Group Meeting on Text Mining (a.k.a. BioLink). June 27, 2003. Brisbane, Australia. [.ps.gz] [.pdf] [.ppt talk]

Andrew I. Schein, Alexandrin Popescul, Lyle H. Ungar, and David M. Pennock. CROC: A New Evaluation Criterion for Recommender Systems. Electronic Commerce Research, Special Issue: World Wide Web Electronic Commerce, Security and Privacy, Editors: Mary Ellen Zurko and Amy Greenwald , Volume 5, Issue 1, January 2005 (pp. 51-74). Draft: [.ps.gz] [.pdf]

Andrew I. Schein, Lawrence K. Saul, and Lyle H. Ungar. A Generalized Linear Model for Principal Component Analysis of Binary Data. Appeared in Proceedings of the 9'th International Workshop on Artificial Intelligence and Statistics. January 3-6, 2003. Key West, FL. This is the paper that gives "Logistic PCA" its name. [talk] [.ps.gz] [.pdf]

Andrew I. Schein, Alexandrin Popescul, Lyle H. Ungar, and David M. Pennock. Methods and Metrics for Cold-Start Recommendations. Appeared in Proceedings of the 25'th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2002), pp 253-260. August 11-15, 2002. Tampere, Finland. [.ps.gz] [.pdf]

Andrew I. Schein, Alexandrin Popescul, Lyle H. Ungar, and David M. Pennock. Generative Models for Cold-Start Recommendations. Workshop on Recommender Systems at SIGIR, 2001. [.ps.gz] [.pdf]

Andrew I. Schein, Alexandrin Popescul, and Lyle H. Ungar. PennAspect: A Two-Way Aspect Model Implementation. University of Pennsylvania Department of Computer and Information Science, Technical Report MS-CIS-01-25. [.ps]

Andrew I. Schein, Jessica C. Kissinger, and Lyle H. Ungar. Chloroplast Transit Peptide Prediction: a Peek Behind the Black Box. Nucleic Acids Research, 2001, Vol 29, No. 16 e82. View Online. A webserver with the software is available. Also, see the software section of my web page for a copy of the code.

Notes

Andrew I. Schein. Notes on the CROC Curve. Unpublished. [.ps.gz] [.pdf]

Andrew Schein. Computation of log(\Phi(z)) For Large Negative z. In 2012, I discovered that the scipy routine scipy.stats.logcdf(z) (normal distribution) produced negative infinite values for moderately negative values of z. I submitted a patch, and now a lot of software relies on the improvement. The linked PDF provides an explanation of the mechanics of the patch and some supporting analysis. The associated issue tracker is here. [.pdf]

Posters

Andrew I. Schein and Lyle H. Ungar. Active Learning for Multi-Class Logistic Regression. LEARNING 2005. Snowbird, Utah. April 5-8, 2005. [.ps.gz][ .pdf ]. Abstract: [.ps.gz][ .pdf ].

Invited Talks

Andrew I. Schein. Active Learning for Logistic Regression. Alberta Ingenuity Centre for Machine Learning, The University of Alberta. January 6, 2005. Talk Overheads [.ps].

Advising

Johnnie F. Caver. Novel Topic Impact on Author Attribution. Masters thesis in Computer Science. 2009. Outstanding thesis award. Co-advised with Craig Martell.
Johnnie's thesis establishes a protocol for measuring author attribution degradation when predicting on novel topics (e.g. document topics not present in training data). Her evaluation task may be reproduced by downloading the NYT corpus, and filtering the corpus using her lists and text extractor.

Software

Below are some of the open source software and patches I have written over the years. The earliest of these date back to sometime around the year 2000. The ordering below is approximately chronological.

PennAspect (link lost to the ages). A java implementation of the Aspect model, a belief network popular in many communities under various names. In the natural language processing and data mining worlds the names "aspect model" or "probabilistic latent semantic indexing" are prevalent terms for this model. A third party has translated our code into C++ and incorporated it into the Lemur toolkit.

Logistic PCA A principal component analysis technique for binary data. Implements the model-fitting strategy introduced in my paper, A generalized linear model for principal component analysis of binary data. The code is implemented as a Matlab procedure.

PCLR An algorithm that predicts protein localization to the chloroplasts in plants. A web-based version of the algorithm described in Nucleic Acids Research, 2001, Vol 29, No. 16 e82. You can download the software that runs on the site.

pa_breakcont Camlp4 3.10 macro for adding break and continue to OCaml loops.

3.10 pa_bounds Camlp4 3.10 compatible release of Martin Jambon's pa_bounds macro. The original is here.

Cython support in Exuberant Ctags (the package continues as Universal Ctags)

usort Optimized C99 routines for sorting numeric data. Much faster than glibc qsort() for this task.

NPSML Naval Postgraduate School Machine Learning Library.

sql.el bug fix for Postgres support in GNU Emacs, credited in the sql.el comments.

scipy.stats.logcdf bug fix described above under "Notes".

cwd_jmp Bookmark relative paths in bash.

golang-gitoperations A Golang library for git automation.

AmIGitEnough A tool for onboarding team members into a rebase-oriented git workflow. I originally developed this inside Amazon where it was very popular at least until the time I exited in 2021. This package needs distribution maintainers for App Image and Brew. Contact me if interested in helping out (or just go for it!).