• Edit
• Help

Staal A. Vinterbo

Staal A. Vinterbo

Dr. Vinterbo received his Ph.D. in computer science from the Norwegian University of Science and Technology (2000). He now works as a Research Scientist at the Decision Systems Group, is an Assistant Professor at Harvard Medical School, and a member of the affiliated faculty of the Harvard-MIT Division of Health Sciences and Technology.

Research Areas

Methods for synthesis and adaptation of predictive models

Theoretically, this work can be said to lie in the intersection of computer science, statistics and medicine, with a focus on machine learning and formal, knowledge based methods. In particular he is concerned with aspects of parsimony in predictive models. Smaller and less complex models are likely to be less costly, both in construction and application, are arguably more robust and applicable, and often exhibit performance not significantly worse than their less parsimonious counterparts.

Methods for disclosure control in disseminated data

Dissemination of medical data in relational form is crucial for biomedical, clinical, bibliographic, administrative, or epidemiological studies. It is also necessary for the development of technologies used in these areas. This work merges logic, mathematics, artificial intelligence, and complexity theory to ensure that confidentiality is maintained in data that is disseminated for research purposes.

One of the common denominators of the research outlined is the application of formal theoretical frameworks.

Contact Information

Location

C1330C, 900 Commonwealth Avenue, Boston, MA 02215

Phone

617-732-7767

Fax

617-525-8804

Email

Misc.

• A guitar chord and scale resolver (written in C) with graphical user interface (written in Tcl/Tk).
• A game for PalmOS(tm) devices.
• Genetic Algorithm for Empirical Variable Selection.
• Estimation of Distribution Algorithm (EDA) library in C++, with an application to the traveling salesman problem (TSP).
• Fuzzy Rules Classifier computation software in R.
• (Fuzzy) Rules Classifier computation software in Python.
• A Python package for reading and writing ARFF format files
• A compact fuzzy inference engine written in R, complete with a "lunar lander control" type simulation example

Publications

Two selected:

• A Stab at Approximating Minimum Subadditive Join.
  • Includes an analysis of the problem for selecting k out of n finite sets such that the cardinality of their union is minimized (minimum coverage).
  • presents both upper and lower approximation bounds for one type of ambiguity based disclosure control schemes (k-ambiguity)
  • is implemented. This python program also implements k-ambiguation by cell suppression.
• Small, fuzzy and interpretable gene expression based classifiers. Deals with constructing classifiers that
  • are human readable
  • can deal with inherent uncertainty
  • offer an abstraction away from the measurement technology used
  • is implemented in R and Python

More of them in a table sortable by columns with search capabilities (javascript) can be found here: Staal's Publications